UNIVERSITY of CALIFORNIA SANTA CRUZ USING HIGH-THROUGHPUT SEQUENCING to STUDY the PROGRESSION of CANCER a Dissertation Submitted

UNIVERSITY OF CALIFORNIA

SANTA CRUZ

USING HIGH-THROUGHPUT SEQUENCING TO STUDY THE PROGRESSION OF CANCER

A dissertation submitted in partial satisfaction of the requirements for the degree of

DOCTOR OF PHILOSOPHY

BIOMOLECULAR ENGINEERING AND BIOINFORMATICS

Mia Griﬀord

December 2014

The Dissertation of Mia Griﬀord is approved:

Professor David Haussler, Chair

Professor Lindsay Hinck

Soﬁe Salama, Ph.D.

Tyrus Miller Vice Provost and Dean of Graduate Studies Copyright c by

Mia Griﬀord

2014 Table of Contents

List of Figures vi

List of Tables viii

Abstract ix

Dedication xi

Acknowledgments xii

1 Introduction 1 1.1 Overview ...... 1 1.2 Background ...... 3 1.2.1 Detecting alterations in cancer with high-throughput technologies 3 1.2.2 Cancer types ...... 5

2 Double Minute Chromosomes in Glioblastoma Multiforme Are Re- vealed by Precise Reconstruction of Oncogenic Amplicons 11 2.1 Introduction ...... 13 2.2 Methods ...... 16 2.2.1 Tumor and normal genome sequencing data ...... 16 2.2.2 Determining breakpoints related to highly ampliﬁed regions . . . 16 2.2.3 Reconstructing amplicons by walking a breakpoint graph . . . . 17 2.2.4 RNA-Seq analysis ...... 19 2.2.5 FISH analysis ...... 19 2.2.6 Exome analysis to estimate prevalence of double minute chromosomes ...... 20 2.2.7 Association of double minute chromosomes/HSR samples with other glioblastoma multiforme tumor features ...... 21 2.3 Results ...... 21 2.3.1 BamBam: a robust method for identifying tumor-speciﬁc variation 21

iii 2.3.2 Reconstruction of candidate double minute chromosomes using whole-genome sequencing ...... 22 2.3.3 Transcriptome data reveals a novel double minute chromosome- associated fusion protein ...... 28 2.3.4 TCGA-06-0648 and TCGA-06-0145 amplicons exist as double minute chromosomes ...... 31 2.3.5 Prevalence of putative double minute chromosomes/HSRs in exome sequencing data ...... 32 2.3.6 Validation of exome sequencing-based prediction of double minute chromosomes/HSR-containing samples ...... 34 2.4 Discussion ...... 36

3 Comprehensive and Integrative Genomic Characterization of Diﬀuse Lower Grade Gliomas 40 3.1 Introduction ...... 41 3.2 Methods ...... 43 3.2.1 Patients ...... 43 3.2.2 Analytic Platforms ...... 43 3.3 Results ...... 44 3.3.1 Histology and molecular subtype ...... 44 3.3.2 Multiplatform integrative analysis ...... 45 3.3.3 Mutational landscape of LGG ...... 49 3.3.4 Signaling networks in LGG ...... 52 3.3.5 IDHwt LGG and GBM ...... 53 3.3.6 Genomic rearrangements and fusion transcripts ...... 55 3.3.7 LGG protein expression ...... 56 3.3.8 From signatures to pathways ...... 57 3.3.9 Clinical characteristics and outcomes of molecular subtypes . . . 59 3.4 Discussion ...... 60

4 Mutation Independent Activation of The Notch Pathway is Associated With Lapatinib Resistance in Her2+ Breast Cancer Cell Lines 66 4.1 Introduction ...... 67 4.2 Methods ...... 69 4.2.1 Cell culture and generation of lapatinib resistant cell lines . . . . 69 4.2.2 Quantitative RT-PCR in drug tolerant persisters ...... 70 4.2.3 Immunoblotting ...... 71 4.2.4 RNA sequencing library preparation ...... 72 4.2.5 RNA sequencing data mapping and analysis ...... 73 4.2.6 Gene expression clustering and pathway enrichment analysis . . 74 4.2.7 Accession codes ...... 74 4.2.8 Mutation calling with RADIA ...... 75 4.3 Results ...... 75

iv 4.3.1 Prolonged incubation in lapatinib robustly generates resistant cell lines ...... 75 4.3.2 Breast cancer DTPs share features with non-small cell lung cancer mutation-independent resistant cells ...... 76 4.3.3 RNA-Sequencing reveals signiﬁcant changes in gene expression . 78 4.3.4 Resistance is mutation independent ...... 80 4.3.5 The Notch pathway is over-expressed in resistant cells ...... 82 4.4 Discussion ...... 85

5 Conclusions 88

A Chapter 3 Supplementary Figures 94

B Chapter 4 Supplementary Figures 98

C List of supplementary data ﬁles 108

Bibliography 110

v List of Figures

1.1 KM Plot of HER2+ vs. HER2- Patients ...... 8 1.2 KM Plot of Patients Grouped by PAM50 Subtype ...... 9

2.1 Reconstruction of TCGA-06-0648 double minutes ...... 26 2.2 Reconstruction of 06-0145 amplicons...... 29 2.3 Novel CPM C-terminal exon expressed from the TCGA-06-0648 double minute chromosome ...... 31 2.4 Visualization of glioblastoma multiforme tumor oncogenic amplicons. . . 32

3.1 Multiplatform analyses point to biologic subtypes deﬁned by IDH mutation and 1p/19q codeletion status...... 48 3.2 Muational landscape and unbiased clustering...... 54 3.3 IDHwt LGG resemble IDHwt GBMs ...... 58 3.4 Integrative analysis and clinical outcome ...... 61

4.1 Lapatinib resistant cell lines were generated by long-term lapatinib treatment...... 77 4.2 Breast cancer DTPs share features with non-small cell lung cancer DTPs. 79 4.3 Notch, PI3K and FOS/JUN pathways are over-expressed in DTEPs compared to untreated cells...... 83 4.4 Notch activation may lead to altered chromatin state...... 87

A.1 The UCSC Cancer Genomics Browser view of GISTIC thresholded copy number calls for the region of chromosome 9 containing CDKN2A/B. . . 95 A.2 Reconstruction of TCGA-CS-5395 circular amplicon ...... 96 A.3 Metrics for cluster of clusters model selection...... 97

B.1 HER2+ lapatinib sensitive breast cancer cell lines used in this study. . . 99 B.2 Drug tolerant expanded persisters begin to grow in 40-70 days...... 100 B.3 DTPs have a distinct chromatin state...... 101 B.4 DE-Seq Finds 6943 Signiﬁcantly Diﬀerentially Expressed Genes. . . . . 102

vi B.5 DE-Seq reveals that untreated cells, DTPs and DTEPs have significantly different expression patterns, with DTEPs being the most distinct. . . . 103 B.6 Pathways significantly over-expressed in DTEPs are highly interconnected.104 B.7 Jag1 and Jag2 are under-expressed in HER2+ breast cancer patients. . 105

vii List of Tables

2.1 RNA-Seqbased expression estimates of 06-0648 double minute chromosome- associated genes ...... 30 2.2 Summary of amplicons found in TCGA GBM exomes ...... 34

3.1 Clinical characteristics of the sample set...... 46

B.1 DE-Seq results conﬁrm that IGF-1R and KDM pathways are over-expressed in resistant cells...... 106 B.2 Genes known to contribute to drug resistance...... 107

viii Abstract

Using high-throughput sequencing to study the progression of cancer

Mia Griﬀord

Cancer is a deadly disease that aﬀects 1 in 2 men and 1 in 3 women in the United

States. The development of high throughput sequencing technologies has allowed for great advancements in cancer therapy, though, we still have a long way to go. Many cancer patients do not receive care soon enough due to late diagnosis, they do not receive the optimum treatment for their speciﬁc tumor type or they receive the optimum treatment, but then develop resistance to it. This thesis will address these three problems and describe how high throughput sequencing technology has allowed us to continue to make advancements in cancer treatment. Speciﬁcally, I will describe one example of how understanding the mechanism of oncogenesis can lead to new options for early diagnosis and monitoring of disease progression in about a quarter of glioblastoma multiforme and IDH wild-type lower grade glioma patients, harboring double minute chromosomes.

I show how comprehensive characterization of lower grade glioma patients has lead to the discovery of a new, more accurate method of subtyping patients, which will allow for patients to receive more appropriate treatment for their speciﬁc tumor type. This is especially important for one subtype of lower grade glioma patients who have poor

ix prognosis and appear very similar to glioblastoma multiforme patients, and therefore should receive similar treatment. I will also describe how development of HER2+ breast cancer cell lines that are resistant to the targeted therapy lapatinib, has led to greater understanding of a mutation-independent mechanism of drug resistance involving activation of the Notch pathway. Patients with a similar mechanism of resistance may beneﬁt from treatment with Notch inhibitors, several of which are currently being tested in clinical trials. All of these ﬁndings display ways in which high throughput sequencing can be used to make advances in our understanding of cancer, ultimately moving towards better treatment for patients and better patient outcomes.

x I dedicate this disseration to my ﬁanc´eAlex, who continuously loved, supported and encouraged me throughout this process.

xi Acknowledgments

I want to thank my committee, Lindsay Hinck, Josh Stuart, Soﬁe Salama and David

Haussler, for their invaluable feedback and support. I especially want to thank Soﬁe

Salama for mentoring me and teaching me so much and David Haussler for his thoughtful insight and encouragement. I want to thank Camilla Forseberg and Scott Boyer for taking me on as a roton in the wetlab, even with no prior experience, and teaching me the fundamentals of wetlab procedures. I also want to thank the many people who guided me and answered all of my questions when I was starting out in the Haussler wet lab, including Bob Sellers, Bryan King, Tracy Ballinger, Dave Greenberg, Courtney

Onodera and Frank Jacobs. I want to thank all of the current and past Haussler wet lab members for their camaraderie, guidance and support, including Olena Morozova,

Andrew Field, Jimi Rosenkrantz, Mari Olsen and Kristoﬀ Tigyi. I want to thank all of the members of the Haussler/Stuart cancer group and the rest of the Haussler dry lab, especially Zack Sanborn, Steve Benz and Sol Katzman, who’ve answered my endless questions and continued to guide me and give me invaluable feedback. I want to thank the countless other graduate students who helped me along this journey including

Lauren Lui, Arjun Rao, Amie Radenbaugh, Sam Ng, John St. John, Phillip Heller and

Wendy Lee. I want to thank all of my teachers and mentors who’ve truly inspired me both at UCSC and at previous colleges, including Lisa Hunter and Anne Metevier of

ISEE, and James Wilson and Heather Shearer in the Writing Program. I lastly want to thank my family and friends who were always there for me and encouraged me to never

xii give up, without them none of this would have been possible.

xiii Chapter 1

Introduction

1.1 Overview

Cancer is the second leading cause of death in the United States [40]. Men have a 1 in

2 risk of developing cancer in their lifetime and women have a 1 in 3 risk [133]. Can- cer treatment has advanced greatly with the invention of high throughput sequencing technology, which has provided great insights into cancer genomes and allowed for the development of targeted therapies which are more eﬀective and have less harmful side eﬀects than traditional chemotherapy. Though several problems with cancer treatment still exist. This thesis will discuss three key problems with cancer treatment and how high throughput sequencing technology has allowed us to make progress in the areas of diagnostics, subtyping and drug resistance.

1 One of the biggest problems with cancer treatment is late diagnosis. Many times patients don’t know they have cancer until it is too late, and in these cases treatment is less eﬀective than if the tumor had been detected earlier. Chapter 2 of this thesis will discuss one feature of gliomas that occur in about a quarter of glioblastoma multiforme patients that may allow for earlier detection and more eﬀective monitoring of disease progression after diagnosis. These features are small circular amplicons called double minute chromosomes, which may be detectable in the blood. The ability to detect and monitor disease progression through the blood would be especially helpful in patients with tumors in places where surgery and biopsies can be risky, like the brain. The computational method for identifying these double minute chromosomes from whole- genome sequencing data will be discussed, as well as a method for predicting the presence and features of double minute chromosomes using exome sequencing data.

Another problem with cancer treatment is determining which treatments are appropriate for each individual. Patients are often categorized into groups, which determine the optimal treatment. Though, proper treatment can only be given if patients are accurately classiﬁed into treatment groups. Chapter 3 will discuss why the current method of categorizing patients in lower grade glioma, by histological type and grade, is not ideal and will propose a new method of subtyping patients that is supported by several independent data types and computational techniques, including high throughput sequencing of both DNA and RNA. This will allow patients to receive the best treatment for their tumor. In particular, this new method of subtyping patients allows for identi-

2 ﬁcation of one group of patients with especially aggressive tumors that behave similarly to glioblastoma multiforme, and should therefore be treated similarly.

When patients are treated with drugs appropriate for their specific tumor type, treatment can be very successful. Often times this involves the use of targeted therapies that directly inhibit proteins critical for tumor growth. These targeted therapies have less side effects than traditional chemotherapy or radiation and can be very effective.

Although, there is one critical problem with targeted therapies. Patients often acquire resistance to these drugs and the therapy becomes ineﬀective. The fourth chapter of this thesis will discuss a mechanism of resistance to the targeted therapy, lapatinib, which was discovered using RNA-sequencing of HER2+ breast cancer cell lines. These cells acquire a mutation independent mechanism of resistance involving activation of the Notch pathway. These ﬁndings could explain what is underlying resistance in many patients and provide new treatment options for these patients.

1.2 Background

1.2.1 Detecting alterations in cancer with high-throughput technolo-

gies

Cancer occurs when normal functioning of a set of genes goes haywire. Alterations in oncogenes can cause cells to divide uncontrollably and altered tumor suppressor genes

3 can prevent cells from undergoing apoptosis, or programmed cell death. There are many diﬀerent types of alterations that can occur in cancer and can either disrupt protein function, cause a protein to signal more than usual or even create novel protein function.

The most well studied genetic alterations are mutations, where a single DNA nucleotide is changed. Alternatively, larger regions of DNA can be either lost (deleted) or duplicated one or more times (ampliﬁed). These changes can be as small as a few bases to as large as an entire chromosome, and are referred to as copy number alterations. The

DNA can also be broken and reconnected incorrectly. Portions of DNA can be incorrectly connected to other regions of DNA or they can be ﬂipped or reversed. We refer to these types of changes as rearrangements or structural variations. These rearrangements can cause two protein coding genes to connect in novel ways, which are referred to as gene fusions. All of these alterations in DNA can be observed with either whole-genome sequencing (sequencing the entire genome) or exome sequencing (sequencing only the portion of genes that code for proteins). Copy number changes can also be observed using a microarray that is specially designed to measure the amount of DNA throughout the entire genome.

DNA can also be altered epigenetically through histone modiﬁcation or methylation.

DNA methylation typically silences the DNA causing genes to no longer be expressed, while histone modiﬁcation can cause either silencing or activation of genes. Histone modiﬁcations can be observed using ChIP-Seq, combining chromatin immunoprecipita-

4 tion with sequencing, and DNA methylation is typically measured using methylation arrays.

Instead of looking at the speciﬁc genetic or epigenetic alterations, we can also observe the eﬀects of these alterations on both RNA production and proteins. Messenger RNA expression can be observed using RNA sequencing and microRNAs can be observed using microRNA sequencing. We can also observe overall amounts of proteins and phosphorylation of proteins using reverse phase protein arrays.

All of these methods allow us to get a more thorough view of what is happening in the cell and help us to determine what is driving cancer progression.

1.2.2 Cancer types

Several diﬀerent high-throughput technologies were used in this thesis to study the progression of both glioma and breast cancer.

1.2.2.1 Glioma

Glioma is a type of tumor that arises from glial cells in the brain. Like many other cancer types, they are identiﬁed by the cell type that the tumor is thought to have arisen from and tumor aggressiveness is labelled with a grade (I - IV). This thesis will discuss several types of gliomas: astrocytomas, which are thought to have arisen from

5 astrocytes, oligodendrogliomas, which are thought to have arisen from oligodendrocytes and oligoastrocytomas, which have a mixed histology with features of both astrocytes and oligodendrocytes. Grade IV astrocytoma is the most aggressive type of glioma and is called glioblastoma multiforme.

1.2.2.2 HER2+ Breast Cancer

Another type of cancer discussed in this thesis is HER2+ breast cancer. Breast cancer patients are routinely tested for several markers and classified into subgroups which help guide treatment. Because there are drugs that specifically target human epidermal growth factor receptor 2 (HER2), and therefore only work for HER2+ patients, breast cancer patients are routinely tested for HER2 amplification and/or over-expression using

ﬂuorescence in situ hybridization (FISH) and/or immunohistochemistry (IHC). They are also routinely tested for the presence of estrogen receptors (ER status) and progesterone receptors (PR status) using nuclear staining. These tests help stratify patients into subtypes, allowing diﬀerent subtypes to receive the type of treatment most appropriate for their cancer.

HER2 is also referred to as HER-2/neu or ERBB2 and is a member of the ErbB or epidermal growth factor receptor (EGFR) protein family [130]. It is a membrane surface- bound receptor tyrosine kinase involved in signal transduction pathways. Expression of

HER2 leads to cell growth and diﬀerentiation. In a normal mammary gland, HER2 is

6 involved in lobuloalveoli development and milk protein secretion [59]. Although HER2 is expressed throughout most developmental stages of the mammary gland, it’s ligands are only expressed during deﬁned stages [28]. HER2 receptors can either homodimerize or heterodimerize with other HER receptors upon ligand binding, and HER2 is the preferred heterodimeric partner for EGFR, HER3 and HER4 [48]. Ampliﬁcation of

HER2 can lead to overexpression, thereby creating an abundance of receptors on the cell surface, which, because of their close proximity, allows for dimerization in the absence of the ligand [147]. This dimerization leads to autophosphorylation and signaling, and, because HER2 is a proto-oncogene, this excess signaling can cause cancer.

HER2 is overexpressed in about 20% of breast cancer patients [106], and these patients have signiﬁcantly worse prognosis than other patients [147]. Figure 1.1 is a Kaplan-

Meyer plot from The Biology Of Cancer textbook showing the proportion of breast cancer patients that have not had a recurrence of cancer over time [147]. The patients are grouped into HER2+ and HER2- groups, and the HER2+ patients have significantly worse prognosis than the rest of the patients. This figure was taken from a 1987 paper by Slamon et. al. [130], and since then other groups have shown similar results using different methods of grouping patients from various datasets [103, 134]. Figure 1.2 is a figure I made, which also shows that HER2 positive patients (as well as luminal B patients) have worse prognosis than other types of breast cancer, when they are grouped into subtypes using a method developed by Parker et. al. [103], called PAM50, which uses the gene expression data from a set of 50 different genes to group the patients into

7 Figure 1.1: Kaplan-Meyer plot showing disease-free survival for HER2+ and HER2- patients. Figure is from The Biology Of Cancer textbook [147]. subtypes. This Kaplan-Meyer plot shows the percentage of patients who have not yet had a distant metastasis, meaning the cancer has not yet spread to another location in the body. I generated this plot using a data set from 683 breast cancer patients composed by Christina Yau [157]. These results make it clear that HER2+ breast cancer is an important type of breast cancer to study.

One drug that is commonly used to treat HER2+ patients and has also been used in previous studies is lapatinib. This thesis will discuss how lapatinib was used to generate resistant cell lines. lapatinib, or Tykerb, is a HER2 kinase inhibitor from

GlaxoSmithKline used to treat advanced or metastatic HER2+ breast cancer. It is usually given in combination with capecitabine, a chemotherapeutic agent, to patients

8 Yau Exp Data − PAM50 gene set subtype prediction

Basal (n=193) 1.0 Her2 (n=101) LumA (n=127) LumB (n=155) Normal (n=107) 0.8 0.6 0.4 % Distant metastasis−free survival 0.2

Cox proportional hazards regression p=3.38e−6 0.0

0 5 10 15 20 25

Figure 1.2: Kaplan-Meyer plot showing the percent of patients that have not yet had a distant metastasis over time. Patients are divided into PAM50 [103] subtypes using gene expression data from 683 breast cancer patients composed by Christina Yau [157].

9 who have already received prior therapy [43]. It crosses the cell membrane and binds to the tyrosine kinase domain of HER2, preventing phosphorylation of HER2 and therefore preventing signal transduction [154].

10 Chapter 2

Double Minute Chromosomes in

Glioblastoma Multiforme Are Revealed by Precise Reconstruction of Oncogenic

Amplicons

As discussed in the introduction, high-throughput sequencing technology has enabled many discoveries in cancer research, helping scientists better understand the mechanisms behind cancer growth. Tools developed by Zachary Sanborn that compute copy number and detect structural variants allowed us to gain insights into the mechanisms driving tumor growth in glioblastoma multiforme (GBM), the most aggressive form of glioma.

One of these mechanisms, which occurs in over 20% of patients, involves oncogenic

11 circular amplicons called double minute chromosomes. Because these double minute chromosomes may be detectable in the blood, they may be used for easier diagnosis and monitoring of disease progression. This chapter describes these amplicons in detail for two GBM patients with whole-genome sequencing data, and describes the estimation of the presence of these amplicons in the larger cohort using exome sequencing data.

This chapter contains text from the paper “Double Minute Chromosomes in Glioblas- toma Multiforme Are Revealed by Precise Reconstruction of Oncogenic Amplicons” by J. Zachary Sanborn, Soﬁe R. Salama, Mia Griﬀord, Cameron W. Brennan, Tom

Mikkelsen, Suresh Jhanwar, Sol Katzman, Lynda Chin and David Haussler that was published in Cancer Research in 2013. Much of this work was completed by Zachary

Sanborn. I contributed to the RNA-Seq analysis, with the help of Sol Katzman in table

2.1 and section 2.2.4. I performed the exome analysis to estimate prevalence of double minute chromosomes and the association of double minute chromosomes/HSR samples with other glioblastoma multiforme tumor features in tables 2.2 and C.1 and sections

2.2.6, 2.3.5 and 2.2.7. I characterized the novel exon, which was discovered by Zachary

Sanborn, in section 2.3.3. I also performed the validation of exome sequencing-based prediction of double minute chromosomes/HSR-containing samples in table C.2 and section 2.3.6.

12 2.1 Introduction

High-throughput methods for whole-genome sequencing have provided researchers with an unprecedented ability to measure the complex state of genomic rearrangements characteristics of most cancers. Numerous methods for inferring structural variation from paired-end sequencing data have been developed [127, 24, 9], but the structural variants called by such methods are often considered only in isolation, and used primarily to identify potential fusion genes. The difficulty in discovering all true structural variants and filtering out false positives makes it hard to use the output of such methods to reassemble large regions of the tumor genome. However, such tumor genome assemblies can reveal the complex structure of the tumor genome and can be used to infer the mechanism by which somatic alterations critical to cancer progression occur, such as amplifications of oncogenes and deletions of tumor suppressors. Ideally, these tumor assemblies will also reveal unique features of the cancer that can be used as diagnostic features to monitor disease progression in the patient.

It is well documented that one mechanism by which genes become highly ampliﬁed in tumors is via the circularization of double-stranded DNA into what are known as double minute chromosomes [8]. Double minute chromosomes have been shown to confer resistance to certain drugs, as well as pass along this resistance nonuniformly to daughter cells. They have been observed up to a few megabases in size, and contain chromatin similar to actual chromosomes, but lack the centromere or telomeres found in normal

13 chromosomes. Because double minute chromosomes lack centromeres, they are, like mitochondria, randomly distributed to daughter cells during cell division [81]. They are generally lost in future generations unless there is some selective pressure to maintain them. For example, when they confer selective growth advantage to tumor cells, they are readily retained at high copy number. In particular, double minute chromosomes containing oncogenes may serve to amplify these genes to hundreds of copies per cell.

Double minute chromosomes are common in several types of cancer, including brain cancers, with an estimated 10% of glioblastoma multiforme tumors bearing double minute chromosomes [11].

Homogeneously staining regions (HSR) represent another common mode of extreme gene ampliﬁcation in cancer, observed in both solid and hematologic cancers [6, 159,

137, 136]. An HSR arises from high copy number tandem duplication of a genomic segment such that it expands the affected chromosome. Because they are embedded within larger chromosomes, fluorescent in situ hybridization (FISH) probes specific to genomic sequences within the HSR give a broad band of staining in a specific position within the larger chromosome, distinct from the focal staining usually observed with a locus specific FISH probe. It is believed that double minute chromosomes and HSRs are related, in that double minute chromosomes can derive from HSRs as well as create HSRs via chromosomal reinsertion [6, 16, 112]. FISH is used to distinguish whether HSRs, double minute chromosomes, or both are present in a given tumor sample. Whatever their form, these highly amplified oncogenes are often key drivers of their respective

14 tumors, and may be vital in their detection, diagnosis, and treatment.

Here, we present methods that analyze high-throughput whole-genome sequencing data from tumor and matched normal samples to detect these extremely ampliﬁed and rearranged regions of the tumor genome. We show that these low-level results can be synthesized to construct accurate local genome assemblies that are circular in nature, suggesting the existence of double minute chromosomes and/or HSRs in the tumor. We use FISH analysis to independently identify double minute chromosomes and HSRs in two tumors samples. In addition, we show that RNA-seq data further supports our circular assemblies, in one case identifying a novel isoform of the gene CPM that co- opts intergenic DNA to create a novel exon. Further analysis of the amplicons is done to distinguish likely driver genes, such as MDM2, from likely passenger genes, such as those disrupted by the amplicon structure. We show that this is made possible by the conﬁguration of the rearranged regions represented by the predicted circular assembly.

Finally, our analysis of a much larger set of samples with whole-exome sequencing data suggests that 20% of glioblastoma multiforme samples harbor highly ampliﬁed oncogenic amplicons.

15 2.2 Methods

2.2.1 Tumor and normal genome sequencing data

Tumor and matched-normal whole-genome BAM ﬁles were downloaded from CGHub

(cghub.ucsc.edu) under the following sample identiﬁers: The Cancer Genome Atlas

(TCGA)-06-0648-01A-01D-0507-08, TCGA-06-0648-10A-01D-0507-08, TCGA-06-0145-

01A-01D-0507-08, and TCGA-06-0145-10A-01D-0507-08. In addition, 264 exome BAM

ﬁles were downloaded from CGHub, with sample identiﬁers given in Supplementary

Table C.1. Whole-genome and exome BAM ﬁles were indexed by samtools [76] and processed by BamBam (See Supplementary Materials and Methods) to determine relative copy number, allele fraction, and structural variants. Supplementary Tables C.1 and S2 list the copy number and rearrangement breakpoints detected by BamBam for exome and whole-genome sequencing data, respectively.

2.2.2 Determining breakpoints related to highly ampliﬁed regions

The read support (number of overlapping reads) for a given breakpoint is directly proportional to the copy number of the regions it connects. Thus, by requiring breakpoints to have a high level of read support, we can filter out breakpoints that are part of a copy-neutral rearrangement, and breakpoints that led to low-copy amplifications and deletions. We can then focus on the breakpoints that are part of highly amplified regions

16 in the tumor. The particular read support threshold is selected such that breakpoints that have read support expected of copy neutral regions of the tumor genome are removed.

2.2.3 Reconstructing amplicons by walking a breakpoint graph

Similar to a recently published method [49], we construct a breakpoint graph by defining a node for each side of an amplified segment of the tumor genome, connecting these two nodes by a segment edge that represents the amplified segment, and defining a set of bond edges, each of which represents a pair of segment sides that are adjacent in the tumor genome. Which amplified segments of the tumor genome are included as segment edges in the breakpoint graph is determined by relative copy number. The bond edges are the highly supported breakpoints found in the manner described earlier.

If an ampliﬁed segment is interrupted by a breakpoint, then that segment will be split into two segment edges.

We visualize the graph with the segments laid out according to genomic position. As- suming all segments have the same relative copy number and each segment side has exactly one bond edge attached to it, we determine the arrangement of the ampliﬁed segments in the tumor genome by starting at the node for the left side of the ﬁrst segment and traversing the segment edge to the node at the other side. The path then follows a bond edge attached to that node over to the segment side to which it is con-

17 nected, then traverses the new segment edge, continuing in this manner of alternatively traversing segment and bond edges until we have returned to the starting node. If there are segment edges remaining that we have not traversed, then we select a new starting node among them and repeat this procedure, until all segment edges are accounted for.

Each cyclic path we determine deﬁnes a separate circular chromosome via the concate- nation of its segments in the direction of traversal. This interpretation of the graph is unambiguous, as the overall direction of traversal of any circular chromosome is imma- terial, as is the order in which we discover them. When the number of bond edges per segment side (i.e., node) in the graph is not uniformly one, or not all segments have the same relative copy number, some segments may require more than one traversal to account for the extra copies, and some walks may terminate at segment sides that have no bond. Whenever multiple walks can be made through the set of bond edges and segments, we enumerate the diﬀerent solutions, including potentially circular solutions that cannot close the loop due to one or more missing bond edges. A single solution that features multiple independent cyclic paths occurs when multiple circular walks are required in the procedure described earlier. Here, distinct sets of bond edges and segments are connected in independent closed loops that lack any bond edges that could fuse the independent loops together into a single, larger closed loop. A toy example breakpoint graph and its solution are described in Supplementary Fig. S1.

When there are multiple solutions, the optimal path(s) through the graph are taken to be those that most closely agree with the observed relative copy number. The number of

18 times a solution traverses a given segment produces an estimate of that segment’s copy number. The root mean square deviation (RMSD) of the segment traversal counts to the observed relative copy number for each solution is calculated, and then the solution(s) with the smallest RMSD value are labeled as optimal.

2.2.4 RNA-Seq analysis

RNA-Seq reads were mapped using BWA [75] and coverage was calculated using BED tools [109]. We calculated the total coverage per transcript per million uniquely mapped read pairs for each gene within the TCGA-06-0648 double minute chromosomes. For

CPM, the coverage over a truncated version of the gene was used as described in the text. These normalized expression metrics of TCGA-06-0648 were compared with a set of nine other TCGA glioblastoma multiforme samples that comprised the original

RNA-Seq cohort for glioblastoma multiforme (Table 2.1).

2.2.5 FISH analysis

FISH was conducted for EGFR/Cep7 and MDM2/Cep12 using prelabeled probes (Vy- sis, Abbott Molecular). Charged slides with 4-mm paraﬃn sections were dewaxed, rinsed, microwaved in 10 mmol/L sodium citrate buﬀer, then digested in pepsin-HCl (40 mg/mL, 10 minutes at 37 ◦C), rinsed, and dehydrated. Probe and slides were codenatured using a HYBrite automated hybridizer at 80 ◦C for eight minutes then hybridized

19 for two to three days at 37 ◦C.

2.2.6 Exome analysis to estimate prevalence of double minute chro-

mosomes

Additional samples that potentially bear double minute chromosomes were identiﬁed in

TCGA glioblastoma multiforme exome datasets as follows. Tumor and matched normal exomes were processed by BamBam to compute relative coverage and identify somatic rearrangements. High copy number peaks were deﬁned as regions with 5-fold increased relative coverage versus their matched normal, a threshold selected to conservatively

ﬁlter out peaks caused by low-level ampliﬁcations and noise (see Supplementary Mate- rials and Methods for details). Glioblastoma multiforme tumors with multiple such high copy number peaks were manually analyzed to discover any oncogenes within peaks, associate peaks with nearby somatic rearrangements, and determine if a sample exhibits multiple peaks with similar copy number levels. Samples were scored as having possible double minute chromosomes if they either contain multiple distinct high copy number peaks with similar copy number levels and at least one peak contained an oncogene, or a single distinct peak containing an oncogene, spanning approximately 1 Mb, and having an associated rearrangement.

20 2.2.7 Association of double minute chromosomes/HSR samples with

other glioblastoma multiforme tumor features

Samples containing likely chromothripsis events were identiﬁed from the set of 26 samples with whole-genome sequencing data determined to have possible double minute chromosome(s) by selecting the subset that had multiple (> 3) high copy number peaks

(18 samples), and were compared with samples containing no high copy number peaks

as determined by whole-exome data (112 samples). There after, t tests were conducted

in R comparing several features between the two groups, including molecular subtype,

survival, and mutation in PTEN, TP53, KEL, IDH1, PIK3R1, PIK3CA, POTEB, NF1,

RB1, and EGFR, ampliﬁcation of PRDM2, MET, MDM2, EGFR, CDK4, CCNE1,

and CCND2, and deletion of RB1, PTEN, PRDM2, MET, and CCND2. Bonferroni-

adjusted P values were reported. TP53 mutations within the two groups were further

evaluated using a two-tailed Fisher’s exact test.

2.3 Results

2.3.1 BamBam: a robust method for identifying tumor-speciﬁc varia-

tion

Considering that a BAM ﬁle storing a single patient’s whole-genome sequence at high

coverage (> 30) can be hundreds of gigabytes in compressed form, a serial analysis of

21 two sequencing datasets requires researchers to store intermediate results that must be merged to conduct a comparative analysis, such as identifying mutations found only in the tumor sample. To overcome this problem, we developed BamBam, a tool that conducts a comparative analysis of a patient’s tumor genome versus his/her germline by simultaneously processing the tumor and matched-normal short-read alignments stored in SAM/BAM-formatted ﬁles [76]. Simultaneous processing of both BAM ﬁles enables

BamBam to eﬃciently calculate tumor relative coverage and allele fraction, discover somatic mutations and germline SNPs, and infer regions of structural variation. The relative coverage and allele fraction estimates made by BamBam can be used to estimate tumor copy number and normal contamination in the sequenced tumor sample (See

Supplementary Methods for details).

2.3.2 Reconstruction of candidate double minute chromosomes using

whole-genome sequencing

We focused on three samples from the initial set of 19 glioblastoma multiforme samples from TCGA subjected to whole-genome sequence analysis where we detected highly ampliﬁed segments overlapping oncogenes, suggestive of double minute chromosomes.

We applied these methods to samples designated TCGA-06-0152, TCGA-06-0648, and

TCGA-06-0145. In each case, whole-genome sequencing of a tumor biopsy was available separately from whole-genome sequencing of a blood sample (matched normal tissue sample). For two of these samples (TCGA-06-0152 and TCGA-06-0648), multiple seg-

22 ments had similar levels of ampliﬁcation whereas the third sample (TCGA-06-0145) had one large ampliﬁed region with further rearrangements internal to the region.

Sample TCGA-06-0648 had a striking pattern of genome amplification in which 16 distinct segments (15 from chromosome 12 and a small fragment from chromosome 9) had similarly high levels of amplification (> 10 copies) and also appeared to be linked to each other by high confidence rearrangement events identified by BamBam (Fig.

2.1A). One of the chromosome 12 segments contains the MDM2 oncogene. Out of the total of 701 putative somatic breakpoints called, 97 breakpoints met the filtering criteria specified in Supplemental Methods. Only 16 of these 97 breakpoints further met or surpassed a chosen minimum read support threshold of 100 to identify breakpoints likely associated with highly amplified regions, including two breakpoints that did not have split read evidence. All of these highly supported breakpoints are proximate to the boundaries of the highly amplified segments clustered on chromosome 12, suggesting that the highly supported breakpoints and the amplifications are directly associated and may represent the rearranged configuration of one (or multiple) amplicons in the tumor’s genome.

Figure 2.1B shows a schematic of the amplified segments of TCGA-06-0648 and their associated breakpoints. This diagram predicts a circular path that completely accounts for all observed breakpoints and amplified segments, resulting in a single 891-kb circular amplicon containing a single copy of MDM2. Because all but two breakpoints were refined by split read solutions, the estimated size of this amplicon is accurate to within

23 approximately 100 bp. This reconstructed circle is consistent with either an array within a larger chromosome of precise tandemly duplicated copies of an initial circular amplicon formed from these 16 genomic segments (assuming single-copy breakpoints joining this tandem array to non-repeated DNA were not suﬃciently covered to be observed) or an extra chromosomal circular DNA (double minute chromosomes; [67]) with average copy number of approximately 84 in the tumor sample. As we can assume neither a perfectly clonal tumor nor a stable number of double minute chromosome copies in every generation of tumor cells, the average copy number of 84 should be considered the number of double minute chromosome copies in the average tumor cell sequenced. We speciﬁcally searched for breakpoints with lower read support within the amplicon region connecting it to other single-copy genomic locations, as these could provide evidence for an insertion site of a tandem array, but found none. Thus, the presence of a double minute chromosome is the more parsimonious explanation of these data, as it does not require us to postulate the existence of one or more pairs of unobserved breakpoints where one or more HSRs each containing multiple exact copies of this 891-kb region are inserted into larger chromosomes.

Double minute chromosomes have been identiﬁed in many tumor types where they often contain oncogenes important for that cancer, such as EGFR in the case of glioblastoma multiforme [146]. The TCGA-06-0648 double minute chromosome contains several protein-coding genes from chromosome 12, including intact copies of the MDM2 oncogene and CAND1, which encodes an inhibitor of cullin ring-ubiquitin ligase com-

24 plexes [36]. In addition, it includes a truncated allele of carboxypeptidase M (CPM), a membrane-bound and secreted protease that cleaves the C-terminal residue of epidermal growth factor (EGF; [35]) . The ampliﬁed allele of CPM lacks the last exon, which should not aﬀect the catalytic or major structural domains of the protein, but removes the amino acids necessary for GPI anchoring of CPM to the plasma membrane and would be expected to result in an exclusively secreted form of CPM [114]. A partial allele of the ras-family gene RAP1B is also present, but as it lacks the promoter and

ﬁrst exon, it is unlikely that this allele is expressed. It seems likely that MDM2 drove the high copy maintenance of this double minute chromosome in TCGA-06-0648, but

CPM and CAND1 could contribute as well.

This region of the tumor genome has all of the hallmarks of a chromothripsis event, suggesting that the double minute chromosome was created by connecting shattered fragments of chromosome 12 and a small region of chromosome 9 into a single circular episome [135]. The observation that multiple regions with uniformly high read depths are connected by a set of structural variants with similarly high-read support suggests that these alterations likely occurred together during a single event, such as chromothripsis, instead of a series of independent focal rearrangements. Furthermore, all breakpoints lack homology or exhibit 2 to 6 bp microhomologies at their junctions, indicating that nonhomologous end-joining (NHEJ) and microhomology-mediated end joining (MMEJ) are the primary DNA double-stranded break-repair mechanisms responsible for constructing the double minute chromosome [88].

25 Sanborn et al.

Figure 1. Reconstruction of TCGA-06-0648 double minutes. A, tumor browser view of a region of chromosome 12 from TCGA-06-0648. The bottom track Figureshows protein-coding 2.1: Reconstruction genes overlapping the ampli offied segments. TCGA-06-0648 The track above shows double relative copy numberminutes. of the tumorA, DNA tumor compared browserwith the normal, showing distinct blocks of elevated copy number with similar total copy number between blocks. The next two tracks show intra- and viewinterchromosomal of a region rearrangement of breakpoints. chromosome All rearrangements 12 from shown are TCGA-06-0648. supported by at least 100 discordant The reads. bottom The type of rearrangementtrack shows is indicated by the color of the line: duplication (red), deletion (blue), inversion (yellow and green), and interchromosomal rearrangement (purple).B,a protein-codingdiagram of the amplified segmentsgenes and overlapping structural variants identi thefied amplified on chromosome 12 segments. and 9 is shown. Walking The through track this diagram above results shows in a circular rel- ativesolution, copy suggesting number the double minute of chromosomesthe tumor diagrammed DNA in C wherecompared segments inverted with relative the to their normal, orientation in theshowing reference genome distinct are indicated by (–) and colored blue. The letters inside the circle correspond to the segments in B. The numbers inside the circle indicate the number of blockssequencing of reads elevated supporting each copy breakpoint. number with similar total copy number between blocks. The next two tracks show intra- and interchromosomal rearrangement breakpoints. All re- arrangementspresent, but as it lacks shown the promoter are and supportedfirst exon, it is unlikely by at leastsuggests 100 that discordant these alterations reads. likely occurred The type together of dur- re- that this allele is expressed. It seems likely that MDM2 drove the ing a single event, such as chromothripsis, instead of a series arrangementhigh copy maintenance is indicated of this double minuteby the chromosome color ofin theof line: independent duplication focal rearrangements. (red), Furthermore, deletion all (blue), break- inversionTCGA-06-0648, (yellow but CPM andCAND1 green),could contribute and interchromosomal as well. points lack homology rearrangement or exhibit2to6bpmicrohomologiesat (purple). B, a di- Thisregionofthe tumorgenome hasallofthe hallmarks of their junctions, indicating that nonhomologous end-joining agrama chromothripsis of the event, amplified suggesting segments that the double and minute structural(NHEJ) variants and microhomology-mediated identified on chromosome end joining (MMEJ) 12 andchromosome 9 is shown. was created Walking by connecting through shattered this fragments diagramare results the primary in DNA a circular double-stranded solution, break-repair suggesting mechan- ofchromosome12 and asmallregionofchromosome9 intoa isms responsible for constructing the double minute chro- thesingle double circular minute episome (22). chromosomes The observation that diagrammed multiple mosome in C where (23). segments inverted relative to theirregions orientation with uniformly inhigh the read referencedepths are connected genome by a are coloredWe applied blue. these same The methods letters to sample inside TCGA-06-0152, the circle set of structural variants with similarly high-read support which also had several highly amplified segments, including correspond to the segments in B. The numbers inside the circle indicate the number of sequencing reads supporting each breakpoint. 6040 Cancer Res; 73(19) October 1, 2013 Cancer Research

26 We applied these same methods to sample TCGA-06-0152, which also had several highly ampliﬁed segments, including segments containing the MDM2, CDK4, and EGFR oncogenes (TCGA; [15]). This analysis predicted two amplicons, in which each amplicon harbored at least one oncogene (Supplementary Fig. S2).

The final case, TCGA-06-0145, exhibits an extreme level of amplification (> 50-fold) of a single approximately 800 kb genomic segment including EGFR that could indicate the presence of an EGFR-double minute chromosome or HSR (Fig. 2.2). In contrast with the other samples, the amplified region of TCGA-06-0145 contains significant variations in the major and minor allele frequency as well as deletion and duplication events with lower read support, which are more compatible with an HSR interpretation. However, again, we were unable to find evidence of breakpoints that would link this amplicon to another genomic region, which argues against an HSR.

The solution to the breakpoint graph of TCGA-06-0145 shows the possibility of three distinct paths that incorporate all breakpoints and explain the observed copy number, and each path predicts a different form of EGFR. EGFR is intact in the dominant path (seven of every nine copies). The remaining two paths, each present in one of nine copies, feature breakpoints that are internal to the EGFR gene, with one path producing a nonfunctional form of EGFR and the other deleting exons 2 to 7 of EGFR. This form of EGFR is known as EGFRvIII, a highly oncogenic, constitutively active form of EGFR that is expressed in multiple tumor types [89]. This is interesting because it suggests two scenarios: (i) EGFRvIII emerges after wild-type EGFR is significantly amplified

27 or (ii) EGFRvIII is created early but cells with more copies of wild-type EGFR have a selective advantage in the tumor population. The former scenario seems most plausible, as the increased EGFRvIII mutant will emerge. Regardless of the true scenario, the ratio of EGFRvIII to wild-type EGFR suggests that a high copy number of oncogenic

EGFRvIII may not be necessary to provide signiﬁcant advantage over the wild-type ampliﬁcation to the growing tumor cell.

2.3.3 Transcriptome data reveals a novel double minute chromosome-

associated fusion protein

For one of the three tumor samples examined in this study (TCGA-06-0648), RNA sequencing was also conducted by TCGA. We analyzed these data together with that from the nine other samples in the initial glioblastoma multiforme RNA-Seq batch to examine the expression of alleles associated with the TCGA-06-0648 double minute chromosome. As expected from the absence of the promoter and ﬁrst exon, RAP1B expression was half that of the other glioblastoma multiforme samples that do not have ampliﬁcations in this region, suggesting that only the intact copy of RAP1B on chromosome 12 is expressed and the double minute chromosome allele is not expressed.

In contrast, MDM2, CAND1, and the ﬁrst eight exons of CPM were expressed at more than 15-fold higher levels than was observed in the glioblastoma multiforme samples lacking ampliﬁcation of these genes (Table 2.1).

28 Double Minute Chromosome Assembly from GBM DNA Sequence Data

segments containing the MDM2, CDK4, and EGFR oncogenes The solution to the breakpoint graph of TCGA-06-0145 (TCGA; ref. 24). This analysis predicted two amplicons, in shows the possibility of three distinct paths that incorporate which each amplicon harbored at least one oncogene (Sup- all breakpoints and explain the observed copy number, and plementary Fig. S2). each path predicts a different form of EGFR. EGFR is intact in The final case, TCGA-06-0145, exhibits an extreme level of the dominant path (seven of every nine copies). The remaining amplification (>50-fold) of a single approximately 800 kb two paths, each present in one of nine copies, feature break- genomic segment including EGFR that could indicate the points that are internal to the EGFR gene, with one path presence of an EGFR-double minute chromosome or HSR producing a nonfunctional form of EGFR and the other (Fig. 2). In contrast with the other samples, the amplified deleting exons 2 to 7 of EGFR. This form of EGFR is known region of TCGA-06-0145 contains significant variations in the as EGFRvIII, a highly oncogenic, constitutively active form of major and minor allele frequency as well as deletion and EGFR that is expressed in multiple tumor types (25). This is duplication events with lower read support, which are more interesting because it suggests two scenarios: (i) EGFRvIII compatible with an HSR interpretation. However, again, we emerges after wild-type EGFR is significantly amplified or (ii) were unable to find evidence of breakpoints that would link EGFRvIII is created early but cells with more copies of wild-type this amplicon to another genomic region, which argues against EGFR have a selective advantage in the tumor population. an HSR. The former scenario seems most plausible, as the increased

Figure 2.2: Reconstruction of 06-0145 amplicons. A, browser shot of the amplicon

on chromosome 7. Tracks as described for Fig. 2.1. The two “Amp. Germline SVs”641 are known germline breakpoints present on the amplicon. B, diagram of the three paths that together form a potential solution of the breakpoint graph for sample TCGA-06- 0145 that accounts for all observed, highly supported breakpoints. The location of the gene EGFR is noted. Note that all paths, A, B, and C, can create circular solutions by using the encompassing tandem duplication shown in red to connect the end of each path to the beginning. The tandem duplication may also connect path A to path B, path C to path B, and so on. C, the eﬀects of each path on the EGFR gene, showing that path B creates an oncogenic form of EGFR, EGFRvIII, through speciﬁc deletion of exons 2 to 7.

29 Sanborn et al.

cannot close the loop due to one or more missing bond edges. A microwaved in 10 mmol/L sodium citrate buffer, then digested single solution that features multiple independent cyclic paths in pepsin–HCl (40 mg/mL, 10 minutes at 37C), rinsed, and occurs when multiple circular walks are required in the dehydrated. Probe and slides were codenatured using a procedure described earlier. Here, distinct sets of bond edges HYBrite automated hybridizer at 80C for eight minutes then and segments are connected in independent closed loops that hybridized for two to three days at 37C. lack any bond edges that could fuse the independent loops together into a single, larger closed loop. A toy example Exome analysis to estimate prevalence of double minute breakpoint graph and its solution are described in Supple- chromosomes mentary Fig. S1. Additional samples that potentially bear double minute When there are multiple solutions, the optimal path(s) chromosomes were identified in TCGA glioblastoma multi- through the graph are taken to be those that most closely forme exome datasets as follows. Tumor and matched normal agree with the observed relative copy number. The number of exomes were processed by BamBam to compute relative times a solution traverses a given segment produces an esti- coverage and identify somatic rearrangements. High copy mate of that segment's copy number. The root mean square number peaks were defined as regions with 5-fold increased deviation (RMSD) of the segment traversal counts to the relative coverage versus their matched normal, a threshold observed relative copy number for each solution is calculated, selected to conservatively filter out peaks caused by low-level and then the solution(s) with the smallest RMSD value are amplifications and noise (see Supplementary Materials and labeled as optimal. Methods for details). Glioblastoma multiforme tumors with multiple such high copy number peaks were manually ana- RNA-Seq analysis lyzed to discover any oncogenes within peaks, associate peaks RNA-Seq reads were mapped using BWA (15) and coverage with nearby somatic rearrangements, and determine if a was calculated using BEDtools (16). We calculated the total sample exhibits multiple peaks with similar copy number coverage per transcript per million uniquely mapped read levels. Samples were scored as having possible double minute pairs for each gene within the TCGA-06-0648 double minute chromosomes if they either contain multiple distinct high copy chromosomes. For CPM, the coverage over a truncated version number peaks with similar copy number levels and at least one of the gene was used as described in the text. These normalized peak contained an oncogene, or a single distinct peak contain- expression metrics of TCGA-06-0648 were compared with a ing an oncogene, spanning approximately 1 Mb, and having an set of nine other TCGA glioblastoma multiforme samples that associated rearrangement. comprised the original RNA-Seq cohort for glioblastoma multiforme (Table 1). Association of double minute chromosomes/HSR samples with other glioblastoma multiforme tumor FISH analysis features FISH was conducted for EGFR/Cep7 and MDM2/Cep12 Samples containing likely chromothripsis events were iden- using prelabeled probes (Vysis, Abbott Molecular). Charged tified from the set of 26 samples with whole-genome sequenc- slides with 4-mm paraffin sections were dewaxed, rinsed, ing data determined to have a possible double minute

Table 1. RNA-Seq–based expression estimates of 06-0648 double minute chromosome–associated genes

Expressiona Sample MDM2 CPM (exons 1–8) CAND1 RAP1B TCGA-GBM-02-0033 2,849 1,180 18,717 22,098 TCGA-GBM-06-0124 3,504 662 15,256 14,654 TCGA-GBM-06-0126 3,247 906 14,808 10,403 TCGA-GBM-06-0155 1,955 911 17,741 32,690 TCGA-GBM-06-0214 5,382 2,783 15,000 15,429 TCGA-GBM-06-0216 3,050 143 20,966 14,595 TCGA-GBM-06-0648 57,126 9,888 480,414 8,493 TCGA-GBM-06-0879 1,235 498 12,833 14,488 TCGA-GBM-12-0692 909 224 14,716 11,022 TCGA-GBM-14-0786 2,305 514 17,484 13,911

Avgb 2,382 630 16,565 16,733 0648 fold increase 23.98 15.70 29.00 0.51

aExpression reported as normalized transcript coverage (coverage per transcript per million uniquely mapped paired-end reads). bExcludes 06-0214 and 06-0648, which have MDM2 ampliﬁcation.

Table 2.1: RNA-Seq-based expression estimates of 06-0648 double minute 6038 chromosome-associatedCancer Res; 73(19) October 1, 2013 genes Cancer Research

For CPM, we observed that many reads originating in exon 8 terminated in a region

1.47 Mb away from it in the normal version of chromosome 12, but only 13.5 kb away

in the double minute chromosome. Closer analysis of this region revealed that the

5’ end of these reads are just downstream of a canonical splice-site acceptor sequence

that generates a new exon encoding a novel 30 amino acid carboxy terminus for the

double minute chromosome-derived CPM allele (Fig. 2.3). This region is not part of

any known transcript and the resulting protein sequence has no strong homology to any

other proteins. This sequence is unlikely to provide a GPI anchor site; therefore, we

anticipate that the double minute chromosome-derived CPM protein would be secreted.

It is not clear what the functional eﬀect of expressing this altered CPM gene would be,

although it may aﬀect EGF metabolism as both membrane bound and secreted forms

of CPM are known to cleave the carboxy-terminal arginine of EGF [86].

In summary, from the point of view of gene expression, the double minute chromosome

30 Sanborn et al.

number of copies subsequently improves the chance that the EGFRvIII mutant will emerge. Regardless of the true scenario, the ratio of EGFRvIII to wild-type EGFR suggests that a high copy number of oncogenic EGFRvIII may not be necessary to provide signiﬁcant advantage over the wild-type ampliﬁcation to the growing tumor cell.

Transcriptome data reveals a novel double minute chromosome-associated fusion protein For one of the three tumor samples examined in this study (TCGA-06-0648), RNA sequencing was also conducted by TCGA. We analyzed these data together with that from the nine other samples in the initial glioblastoma multiforme RNA- Figure 4. Visualization of glioblastoma multiforme tumor oncogenic amplicons. FISH analysis of formalin-fixed paraffin-embedded blocks Seq batch to examine the expression of alleles associated with matching samples 06-0145 (A) and 06-0648 (B). Probes for the the TCGA-06-0648 double minute chromosome. As expected centromeric region of chromosome 7 (A) and chromosome 12 (B) shown from the absence of the promoter and first exon, RAP1B ex- in green gave two to four foci per cell. Probes for EGFR (A) and MDM2 (B) pression was half that of the other glioblastoma multiforme are shown in pink. White arrows point to broad, HSR-like staining samples that do not have amplifications in this region, sug- patterns, and yellow arrows point to discrete extrachromosomal spots. gesting that only the intact copy of RAP1B on chromosome 12 is expressed and the double minute chromosome allele is teins. This sequence is unlikely to provide a GPI anchor site; not expressed. In contrast, MDM2, CAND1, and the first eight therefore, we anticipate that the double minute chromosome- exons of CPM were expressed at more than 15-fold higher levels derived CPM protein would be secreted. It is not clear what the than was observed in the glioblastoma multiforme samples functional effect of expressing this altered CPM gene would be, lacking amplification of these genes (Table 1). although it may affect EGF metabolism as both membrane- For CPM, we observed that many reads originating in exon 8 bound and secreted forms of CPM are known to cleave the terminated in a region 1.47 Mb away from it in the normal carboxy-terminal arginine of EGF (26). version of chromosome 12, but only 13.5 kb away in the double In summary, from the point of view of gene expression, the minute chromosome. Closer analysis of this region revealed double minute chromosome results in overexpression of that the 50 end of these reads are just downstream of a MDM2, CAND1, and a novel form of CPM in this tumor sample. canonical splice-site acceptor sequence that generates a new exon encoding a novel 30 amino acid carboxy terminus for the TCGA-06-0648 and TCGA-06-0145 amplicons exist as double minute chromosome-derived CPM allele (Fig. 3). This double minute chromosomes region is not part of any known transcript and the resulting The ability to distinguish HSRs from double minute chro- protein sequence has no strong homology to any other pro- mosomes from short-read sequencing data is limited. To independently assess the nature of the amplification events in these tumors, we conducted FISH analysis on paraffin sections derived from tumors TCGA-06-0648 and TCGA-06- 0145 using probes to MDM2 (amplified in TCGA-06-0648) and EGFR (amplified in TCGA-06-0145; Fig. 4). Material was unavailable for TCGA-06-0152. In 0648, the MDM2 probe gave a punctate pattern throughout the nucleus, indicating many nonchromosomal sites of MDM2, typical of double minute chromosomes. In contrast, the EGFR probe gave a broad pattern of staining in addition to punctate spots for TCGA- 06-0145, consistent with a combination HSR and double minute chromosomes.

Prevalence of putative double minute chromosomes/ HSRs in exome sequencing data Exome sequencing is cheaper than whole-genome sequencing, and samples of tumors that have been subjected to exome sequencing are more plentiful. Evidence for double minute chromosomes can be obtained from exome-sequencing by searching for patterns of multiple regions of high-level ampli- Figure 2.3: Novel CPMFigure 3. Novel C-terminalCPM C-terminal exon exon expressed expressed from the TCGA-06-0648 from the TCGA-06-0648ﬁcation overlapping at least one oncogene, a pattern common double minute chromosome. The top line shows the appropriate region to the double minute chromosome-containing samples we double minute chromosome.from Fig. 1C, with sequencingThe top reads line where shows the other the pair maps appropriate to region from Fig. 2.1C, with sequencingexon reads 8 of CPM wherebelow. The the bottom other panel showspair the maps orientation to exonof the new 8 of CPManalyzed. below. In contrast, The broad or chromosome arm-level ampli- exon relative to CPM and its amino acid sequence. ﬁcation events as well as focal events with a modest level of bottom panel shows the orientation of the new exon relative to CPM and its amino acid sequence. 6042 Cancer Res; 73(19) October 1, 2013 Cancer Research results in overexpression of MDM2, CAND1, and a novel form of CPM in this tumor sample.

2.3.4 TCGA-06-0648 and TCGA-06-0145 amplicons exist as double

minute chromosomes

The ability to distinguish HSRs from double minute chromosomes from short-read sequencing data is limited. To independently assess the nature of the ampliﬁcation events in these tumors, we conducted FISH analysis on paraﬃn sections derived from tumors

TCGA-06-0648 and TCGA-06-0145 using probes to MDM2 (ampliﬁed in TCGA-06-

0648) and EGFR (ampliﬁed in TCGA-06-0145; Fig. 2.4). Material was unavailable

31 Sanborn et al.

Transcriptome data reveals a novel double minute chromosome-associated fusion protein For one of the three tumor samples examined in this study (TCGA-06-0648), RNA sequencing was also conducted by TCGA. We analyzed these data together with that fromFigure the 2.4: Visualization of glioblastoma multiforme tumor oncogenic am- Figure 4. Visualization of glioblastoma multiforme tumor oncogenic nine other samples in the initial glioblastoma multiformeplicons. RNA- FISH analysis of formalin-fixed paraffin-embedded blocks matching samples amplicons. FISH analysis of formalin-fixed paraffin-embedded blocks Seq batch to examine the expression of alleles associated06-0145 with (A)matching and 06-0648 samples (B). 06-0145 Probes (A) for and the 06-0648 centromeric (B). Probes region for the of chromosome 7 (A) the TCGA-06-0648 double minute chromosome. As expectedand chromosomecentromeric 12 (B) region shown of chromosomein green gave 7 (A) two and to chromosome four foci per 12 (B)cell. shown Probes for EGFR from the absence of the promoter and first exon, RAP1B(A)ex- and MDM2in green (B) gave are showntwo to four in focipink. per White cell. Probes arrows for EGFR point(A) to and broad,MDM2 HSR-like(B) staining pression was half that of the other glioblastoma multiformepatterns, andare yellow shown arrows in pink. point White arrowsto discrete point to extrachromosomal broad, HSR-like staining spots. samples that do not have amplifications in this region, sug- patterns, and yellow arrows point to discrete extrachromosomal spots. gesting that only the intact copy of on chromosome RAP1B for TCGA-06-0152. In 0648, the MDM2 probe gave a punctate pattern throughout 12 is expressed and the double minute chromosome allele is teins. This sequence is unlikely to provide a GPI anchor site; fi not expressed. In contrast, MDM2, CAND1, and the rstthe eight nucleus, indicatingtherefore, we many anticipate nonchromosomal that the double sites of minute MDM2, chromosome- typical of double minute exons of CPM were expressed at more than 15-fold higher levels derived CPM protein would be secreted. It is not clear what the than was observed in the glioblastoma multiforme sampleschromosomes.functional In contrast, effect the of EGFR expressing probe this gave altered a broadCPM patterngene would of staining be, in addition lacking amplification of these genes (Table 1). although it may affect EGF metabolism as both membrane- For CPM, we observed that many reads originating in exonto punctate 8 spotsbound for and TCGA-06-0145, secreted forms consistent of CPM are with known a combination to cleave HSR the and double terminated in a region 1.47 Mb away from it in the normal carboxy-terminal arginine of EGF (26). version of chromosome 12, but only 13.5 kb away in the doubleminute chromosomes.In summary, from the point of view of gene expression, the minute chromosome. Closer analysis of this region revealed double minute chromosome results in overexpression of that the 50 end of these reads are just downstream of a MDM2, CAND1, and a novel form of CPM in this tumor sample. canonical splice-site acceptor sequence that generates a new exon encoding a novel 30 amino acid carboxy terminus for2.3.5 the PrevalenceTCGA-06-0648 of putative and TCGA-06-0145 double minute amplicons chromosomes/HSRs exist as in ex- double minute chromosome-derived CPM allele (Fig. 3). This double minute chromosomes region is not part of any known transcript and the resulting ome sequencingThe ability to data distinguish HSRs from double minute chro- protein sequence has no strong homology to any other pro- mosomes from short-read sequencing data is limited. To independently assess the nature of the amplification events fi Exome sequencingin these is cheaper tumors, than we conducted whole-genome FISH sequencing, analysis on and paraf samplesn of tumors sections derived from tumors TCGA-06-0648 and TCGA-06- fi that have been0145 subjected using probes to exome to MDM2 sequencing(ampli areed more in TCGA-06-0648) plentiful. Evidence and for double EGFR (amplified in TCGA-06-0145; Fig. 4). Material was minute chromosomesunavailable can for be TCGA-06-0152. obtained from exome-sequencing In 0648, the MDM2 byprobe searching gave for patterns a punctate pattern throughout the nucleus, indicating many nonchromosomal sites of MDM2, typical of double minute chromosomes. In contrast, the EGFR probe gave a broad pattern of staining in addition32 to punctate spots for TCGA- 06-0145, consistent with a combination HSR and double minute chromosomes.

6042 Cancer Res; 73(19) October 1, 2013 Cancer Research of multiple regions of high-level amplification overlapping at least one oncogene, a pattern common to the double minute chromosome-containing samples we analyzed. In contrast, broad or chromosome arm-level amplification events as well as focal events with a modest level of amplification (e.g., duplications) are not expected to be part of a double minute chromosome. To estimate the prevalence of double minute chromosomes/HSRs in glioblastoma multiforme, we searched a set of 264 TCGA samples with tumor and matched-normal exome sequencing data for signatures of double minute chromosomes. As detailed in the Materials and Methods, we first conducted a computational survey of samples to identify samples exhibiting focal extreme amplification(s), prioritizing those samples with multiple distinct peaks on one or more chromosomes. A careful manual review of these samples was conducted to assess the likelihood that the sample contains a double minute chromosome, looking for the quality of relative copy number calls, any evidence of structural variation associated with the amplified peaks, and the presence of potential oncogenes.

As described in Table 2.2 and Supplementary Table C.1, 61 samples (23%) have features suggesting the presence of a double minute chromosome. A total of 121 oncogenes were amplified across these 61 samples, with at least one oncogene identified in every putative double minute chromosome. EGFR was the oncogene most frequently associated with these high-level amplicons, followed by CDK4 and MDM2. MYCN, which has been identified in a number of glioblastoma multiforme double minute chromosomes [11], was associated with amplicons in two samples. As a group, the putative double minute

33 Double Minute Chromosome Assembly from GBM DNA Sequence Data

amplification (e.g., duplications) are not expected to be part of cons for these samples, although the possibility remains that a double minute chromosome. To estimate the prevalence of rearrangement breakpoints allowing for a circular solution were double minute chromosomes/HSRs in glioblastoma multi- not detected for technical reasons. For 2 of the 25 samples where forme, we searched a set of 264 TCGA samples with tumor we did not predict a double minute chromosome/HSR based on and matched-normal exome sequencing data for signatures of the exome sequencing data, we also did not detect highly double minute chromosomes. As detailed in the Materials and amplified genomic regions with associated rearrangements in Methods, we first conducted a computational survey of sam- the whole-genome sequencing data. ples to identify samples exhibiting focal extreme amplification This larger set of samples with circular amplicons allowed us (s), prioritizing those samples with multiple distinct peaks on to look for common features associated with these samples one or more chromosomes. A careful manual review of these compared with TCGA glioblastoma multiforme samples where samples was conducted to assess the likelihood that the exome data suggests no amplicons. Previous studies identify- sample contains a double minute chromosome, looking for ing medulloblastomas with chromothripsis-associated double the quality of relative copy number calls, any evidence of minute chromosomes and other complex genetic rearrange- structural variation associated with the amplified peaks, and ments noted that such samples harbor p53 mutations (27, 28). the presence of potential oncogenes. However, there was no enrichment of p53 mutations in the As described in Table 2 and Supplementary Table S1, 61 samples with circular amplicons in this sample cohort. The samples (23%) have features, suggesting the presence of a only significant association we observed was with PTEN dele- double minute chromosome. A total of 121 oncogenes were tion (Bonferroni-adjusted P value 0.0052), a common event ¼ amplified across these 61 samples, with at least one oncogene in glioblastoma multiforme tumors. identified in every putative double minute chromosome. EGFR Focusing on the 16 samples where we successfully recon- was the oncogene most frequently associated with these high- structed circular amplicons with at least one oncogene, a level amplicons, followed by CDK4 and MDM2. MYCN, which mixture of simple and complex amplicons was observed. One has been identified in a number of glioblastoma multiforme- sample harbored two simple circular amplicons, each contain- double minute chromosomes(6), was associated with ampli- ing one genomic segment with an oncogene and the remaining cons in two samples. As a group, the putative double minute five had a single oncogenic amplicon containing one genomic chromosomes/HSR-containing samples showed similar sur- segment (Supplementary Table S3). Ten samples are complex, vival to the cohort of samples of analyzed by exome sequencing harboring multiple highly amplified segments from one or (data not shown). Taken together, these results suggest that more chromosomes. Of these 10, six samples have at least two almost a quarter of glioblastoma multiforme samples have circular amplicons. Together, these results strengthen our oncogenic amplicons present at high copy number. estimates of the prevalence double minute chromosomes/HSR amplicons in glioblastoma multiformes and suggest that sam- Validation of exome sequencing-based prediction of ples containing such amplicons can often harbor multiple double minute chromosomes/HSR-containing samples independent amplicons. Recently, the TCGA project conducted whole-genome sequencing on an additional 25 tumor/normal pairs from the glioblastoma multiforme cohort. This new dataset contains 23 Discussion samples that we predicted to harbor a double minute chromo- The ability to integrate relative copy number with break- some/HSR on the basis of our curation of the exome data for points enables us to understand the topology of vital parts of these samples. Although complete analysis of this data is being the cancer genome. By examining the overall pattern of ampli- pursued by the Analysis Working Group, we conducted a fication events in the TCGA glioblastoma multiforme whole- preliminary analysis using BamBam and the methods described genome sequencing data, we found that some samples had earlier to identify circular amplicons in these samples (sum- multiple highly amplified segments of similar copy number. marized in Supplementary Table S3). For 16 of the 23 samples, Surprisingly, for many of these highly amplified tumor regions, we were able to reconstruct at least one circular amplicon. For we are able to completely explain both the observed copy the remaining 7 of the 23 samples, multiple highly amplified number and highly supported breakpoints surrounding them peaks were identified with breakpoints connecting many, but by solving a simple breakpoint graph, which describes the not all, of the peaks. We could not reconstruct circular ampli- order and orientation of the highly amplified segments in the tumor genome. For the three glioblastoma multiforme samples analyzed in detail here, the optimal solutions to the breakpoint Table 2. Summary of amplicons found in TCGA- graphs of amplified segments are circular amplicons. These GBM exomes circular solutions suggest that the observed amplified regions are double minute chromosomes or HSRs. Sequence coverage of 30 could be limiting our ability to Total samples 264 Â Total potential double minutes 61 detect breakpoints associated with the chromosomal integra- Potential DM with EGFR 46 tion site of an array of these amplicons, as the copy number of Potential DM with CDK4 18 the breakpoint at the integration site is much lower than that Potential DM with MDM2 14 of the amplicon. Thus, at this coverage we cannot reliably distinguish between a double minute chromosome and an

Table 2.2: Summarywww.aacrjournals.org of amplicons found in TCGA GBM exomes Cancer Res; 73(19) October 1, 2013 6043 chromosomes/HSR-containing samples showed similar survival to the cohort of samples analyzed by exome sequencing (data not shown). Taken together, these results suggest that almost a quarter of glioblastoma multiforme samples have oncogenic amplicons present at high copy number.

2.3.6 Validation of exome sequencing-based prediction of double minute

chromosomes/HSR-containing samples

Recently, the TCGA project conducted whole-genome sequencing on an additional 25 tumor/normal pairs from the glioblastoma multiforme cohort. This new dataset contains 23 samples that we predicted to harbor a double minute chromosome/HSR on the basis of our curation of the exome data for these samples. Although complete analysis of this data is being pursued by the Analysis Working Group, we conducted a preliminary analysis using BamBam and the methods described earlier to identify circular amplicons in these samples (summarized in Supplementary Table C.2). For 16 of the 23 samples, we were able to reconstruct at least one circular amplicon. For the remaining 7 of the

23 samples, multiple highly ampliﬁed peaks were identiﬁed with breakpoints connecting

34 many, but not all, of the peaks. We could not reconstruct circular amplicons for these samples, although the possibility remains that rearrangement breakpoints allowing for a circular solution were not detected for technical reasons. For 2 of the 25 samples where we did not predict a double minute chromosome/HSR based on the exome sequencing data, we also did not detect highly ampliﬁed genomic regions with associated rearrangements in the whole-genome sequencing data.

This larger set of samples with circular amplicons allowed us to look for common features associated with these samples compared with TCGA glioblastoma multiforme samples where exome data suggests no amplicons. Previous studies identifying medulloblastomas with chromothripsis-associated double minute chromosomes and other complex genetic rearrangements noted that such samples harbor p53 mutations [111, 94]. However, there was no enrichment of p53 mutations in the samples with circular amplicons in this sample cohort. The only signiﬁcant association we observed was with PTEN deletion

(Bonferroni-adjusted P value = 0.0052), a common event in glioblastoma multiforme tumors.

Focusing on the 16 samples where we successfully reconstructed circular amplicons with at least one oncogene, a mixture of simple and complex amplicons was observed. One sample harbored two simple circular amplicons, each containing one genomic segment with an oncogene and the remaining ﬁve had a single oncogenic amplicon containing one genomic segment (Supplementary Table C.2). Ten samples are complex, harboring multiple highly ampliﬁed segments from one or more chromosomes. Of these 10, six

35 samples have at least two circular amplicons. Together, these results strengthen our estimates of the prevalence double minute chromosomes/HSR amplicons in glioblastoma multiformes and suggest that samples containing such amplicons can often harbor multiple independent amplicons.

2.4 Discussion

The ability to integrate relative copy number with breakpoints enables us to understand the topology of vital parts of the cancer genome. By examining the overall pattern of amplification events in the TCGA glioblastoma multiforme whole-genome sequencing data, we found that some samples had multiple highly amplified segments of similar copy number. Surprisingly, for many of these highly amplified tumor regions, we are able to completely explain both the observed copy number and highly supported breakpoints surrounding them by solving a simple breakpoint graph, which describes the order and orientation of the highly amplified segments in the tumor genome. For the three glioblastoma multiforme samples analyzed in detail here, the optimal solutions to the breakpoint graphs of amplified segments are circular amplicons. These circular solutions suggest that the observed amplified regions are double minute chromosomes or HSRs.

Sequence coverage of 30X could be limiting our ability to detect breakpoints associated with the chromosomal integration site of an array of these amplicons, as the copy number

36 of the breakpoint at the integration site is much lower than that of the amplicon. Thus, at this coverage we cannot reliably distinguish between a double minute chromosome and an HSR with our bioinformatic analysis. The availability of tissue sections derived from these same tumor samples did allow us to directly address this issue for two samples. FISH analysis of one sample, TCGA-06-0648, is consistent with a double minute chromosome, whereas another, TCGA-06-0145, gives a pattern suggestive of a combination of double minute chromosome and HSR.

Much can be learned through precise knowledge of the amplicon structure. Genes whose coding or promoter regions are disrupted by the amplicon structure are obvious passenger gene candidates, provided that transcriptional machinery is unable to create a novel transcript like the one observed with the birth of a new exon for the CPM gene. Highly expressed genes such as MDM2 that remain intact may drive tumor development and/or play a role in tumorigenesis, using the double minute chromosome’s ability to rapidly reproduce to signiﬁcantly increase the oncogenic capacity of these cells.

The highly ampliﬁed state of these oncogene-harboring amplicons indicates that they have strong oncogenic potential and, thus, confer a selective advantage to the tumor cell. Their formation was likely a key event in the tumorigenesis of these glioblastoma multiforme tumors and they are likely to persist overtime.

Examining the larger set of over 250 TCGA glioblastoma multiforme samples for which there is exome data suggests that oncogenic amplicons are quite prevalent as they are found in almost a quarter of the samples. Using recently generated whole-genome

37 sequencing data for an additional 23 samples from the set of samples where we had predicted double minute chromosomes/HSRs from exome sequencing data, we were able to confirm all of these samples had highly amplified genomic segments with rearrangements connecting at least some of the amplified segments. Further, we were able to reconstruct circular amplicons for 16 of these samples and discovered that six had multiple amplicons as we had observed for TCGA-06-0152, suggesting that the presence of multiple amplicons is common, and that they persist by conferring a combined selective advantage to the tumor cells. These results bolster the notion that chromothripsis type events occur with reasonable frequency in glioblastoma multiforme and, through amplification of oncogene expression, contribute to tumorigenesis.

The prevalence of these amplicons suggests that tumor specific breakpoints associated with double minute chromosomes amplicons may be a potential diagnostic for the presence of tumor cells in a significant fraction of patients with glioblastoma multiforme, especially if double minute chromosome-derived DNA is present in the blood. Several recent observations suggest the possibility of finding such tumor DNA in the blood of patients who have glioma. Skog and colleagues reported microvesicles that bud off from glioblastoma multiforme tumor cells with lipid bilayers intact, carrying cytoplas- mic contents of the tumor cell such as mRNA, miRNA, and angiogenic proteins [129].

These vesicles can deliver their contents to other cells or blood. More recently, Balaj and colleagues have isolated microvesicles containing single stranded DNA (ssDNA) with ampliﬁed oncogenic sequences, in particular c-Myc [7]. Tumor nucleic acids leaked

38 into the blood in this manner or via macrophages that engulf necroticor apoptotic cells have been proposed as possible disease biomarkers in several cancers including glioma

[124, 69]. Indeed, a recent study robustly detected tumor-associated rearrangements from patients who have breast and colorectal cancer via high-throughput sequencing of the cell-free, plasma fraction of blood [71]. The methods described in this study could readily be applied to such sequencing data and, thus, potentially provide a noninvasive method to characterize and monitor patients with glioma.

39 Chapter 3

Comprehensive and Integrative

Genomic Characterization of Diﬀuse

Lower Grade Gliomas

Another form of gliomas, termed lower-grade gliomas (LGG), have historically been thought of as less aggressive than GBM and often eventually progress to GBM. This chapter describes work by The Cancer Genome Atlas (TCGA) consortium, for which I have been heavily involved, describing a new method of classiﬁcation of LGG samples.

This classification reveals one subtype of LGG which is just as aggressive as GBM and has features very similar to GBM, suggesting these patients may benefit from more aggressive treatment. This work demonstrates how this new method of clinical classification is more accurate and groups patients into more clinically relevant subtypes than

40 the current standard classiﬁcation, which is histology and grade.

This chapter contains text from the paper ”Comprehensive and Integrative Genomic

Characterization of Diﬀuse Lower Grade Gliomas” by The Cancer Genome Atlas Net- work submitted to The New England Journal Of Medicine in 2014. I performed the cluster of clusters analysis described in ﬁgures 3.1 and A.3 and sections 3.2.2, 3.3.2 and

3.3.8, the comparison of copy number alterations between IDHwt LGG and GBM in

ﬁgure 3.3A and sections 3.3.3, 3.3.5 and 3.3.8 and processed data used to compare mutations, copy number changes and structural variants in ﬁgure 3.3B and sections 3.3.3,

3.3.5 and 3.3.8 (the generation of the figure was completed by Olena Morozova). I also analyzed deletions in CDKN2A/B described in figure A.1 and section 3.3.8. I processed and analyzed genomic rearrangement data from the tool, BamBam (developed and run by Zachary Sanborn), described in figure 3.3B, table C.3 and section 3.3.6 and performed the double minute/breakpoint-enriched region analysis described in figures 3.3A and A.2, table C.3 and section 3.3.6. Finally, I performed the initial survival analysis described in figure 3.4B and section 3.3.9, although the final version of the figure was re-made by Laila Poisson.

3.1 Introduction

Diﬀuse low grade and intermediate grade gliomas (WHO grades II and III, hereafter called lower grade gliomas (LGG)) are inﬁltrative neoplasms that arise most often in

41 the cerebral hemispheres of adults and include astrocytomas, oligodendrogliomas and oligoastrocytomas of World Health Organization (WHO) grades II and III [78, 98]. Due to their highly invasive nature, complete neurosurgical resection is nearly impossible, and residual neoplasms eventually undergo malignant progression, yet with highly variable intervals. A subset will progress to glioblastoma (GBM, WHO grade IV) within months, while others remain stable for years. Similarly, survival ranges widely from 1-15 years, with some LGG showing impressive chemosensitivity [18, 82]. Current treatment varies with extent of resection, histology, grade and results from ancillary testing, and includes clinical monitoring, chemotherapy, and radiation therapy, with salvage options available upon treatment failure [14, 120]. Although comprehensive molecular proﬁling has greatly clariﬁed the pathogenic landscape of GBM, LGG remain less understood

[15, 21].

While histopathologic classification of LGG is time-honored, it suffers from high intra- and interobserver variability [78, 31, 143]. Consequently, clinicians increasingly rely on genetic classifications to guide clinical decision-making [120, 139, 140, 148]. Mutations in IDH1 and IDH2 characterize the majority of LGG and define a subtype with favorable prognosis [57, 105, 155]. Those with chromosome 1p/19q co-deletion, most often noted in oligodendrogliomas, have better responses to chemotherapy and longer survival [17,

19, 46]. TP53 and ATRX mutations are more frequent in astrocytomas and are also important markers of clinical behavior [61]. To gain additional insights, we performed a comprehensive analysis of 293 LGG, using multiple advanced molecular platforms, to

42 provide a deeper understanding of their features, to classify them in a clinically relevant manner, and to shed light on potential targets for therapy.

3.2 Methods

3.2.1 Patients

Tumor samples were from 293 adults with previously untreated LGG (WHO grade II and

III) and included 100 astrocytomas, 77 oligoastrocytomas and 116 oligodendrogliomas

[149]. Diagnoses were established at contributing institutions and a quality review was performed by TCGA neuropathologists. Patient characteristics are described in Tables

3.1, S1 and S2, and in Supplementary Information. We obtained appropriate consent from relevant Institutional Review Boards. The range of patient ages, tumor locations, clinical histories and outcomes, histologies and grades were typical of patients diagnosed with diﬀuse glioma [149].

3.2.2 Analytic Platforms

We performed exome sequencing (n= 289), DNA copy number profiling (n=285), mRNA sequencing (n=277), microRNA sequencing (n=293), DNA methylation profiling (n=289), and reverse-phase protein lysate array (RPPA) protein expression profiling (n=255).

Complete data for all platforms were available for a core set of 254 LGG. Whole genome

43 sequencing and low pass whole genome sequencing were performed on 21 and 52 samples, respectively. The molecular data package associated with this report was frozen on

1/31/2014. Clinical data were frozen on 8/25/2014. The complete list of data sets is provided in Table S1. Data are available through the Cancer Genome Atlas (TCGA) data portal (https://tcga-data.nci.nih.gov/tcga) and CGHub (https://cghub.ucsc.edu/). In addition to single platform molecular characterization, we also performed unsupervised analysis that integrated results from multiple platforms, including Cluster of Cluster

(CoC) and OncoSign [27]. Brieﬂy, CoC is a second level clustering of class assignments derived from each individual molecular platform. OncoSign classiﬁes tumor samples based on similarity of recurrent somatic events. Statistical analysis includes Fishers exact test for association between categorical variables; one-way ANOVA for association with continuous outcomes; Kaplan-Meier estimates of survival with log-rank tests between strata; and Cox proportional-hazards regression for multi-predictor models of survival. In depth descriptions of methods are found in the Supplemental text.

3.3 Results

3.3.1 Histology and molecular subtype

In order to compare tumor profiles derived from molecular platforms with both histologic classification and with markers frequently used in clinical practice (IDH mutation and 1p/19q co-deletion status), we classified all LGG in our cohort as either:

44 1) IDH-mutated, 1p/19q co-deleted (IDHmut-codel); 2) IDH-mutated, 1p/19q non- co-deleted (IDHmut-non-codel); or 3) IDHwildtype (IDHwt). Consistent with prior reports [19, 2, 44, 85], we showed a strong correlation of IDHmut-codel status with oligodendroglioma histology (69/84) (Table 3.1). IDHmut-non-codel samples (n=139;

50% of cohort) represented a mixture of LGG histologies, enriched for astrocytomas and oligoastocytomas. IDHwt samples were mostly astrocytomas (31/55) and grade III

LGG (42/55), yet also included other histologies and grades (Table 3.1). Overall, IDH-

1p/19q status correlated well with oligodendrolioma classiﬁcation, but only modestly with astrocytoma and oligoastrocytoma histologies.

3.3.2 Multiplatform integrative analysis

Grouping of tumors based on molecular patterns has shown robust associations with disease biology in GBM [95, 107, 145]. To assess whether similar associations could be established in LGG, we performed unsupervised clustering of molecular proﬁles derived from four independent platforms and demonstrated well-deﬁned subtypes based on DNA methylation (5 clusters; Figure S1-5); gene expression (4 clusters; Figure S6-7); DNA copy number (3 clusters; Figure S8); and microRNA (4 clusters; Figure S9-10).

To integrate data from individual molecular platforms and compare proﬁles to histology and IDH-1p/19q status, cluster group assignments deﬁned by the four individual platforms (DNA methylation, mRNA, DNA copy number, and microRNA) were used

45 Table 1: Clinical characteristics of the sample set. Clinical characteristics are presented in total and by IDH mutant status and 1p/19q co-deletion status within the IDH mutant group. The number of samples with known information is given when complete data are not available. Count, and percentage within group, is given for categorical variables and distributions are compared by Fisher’s exact test. Mean, SD, and range are given for age at diagnosis and ANOVA was used to compare groups. Significance is noted as ư 0.10

Total IDHmut IDHmut IDHwt (n=278£) codel non-codel (N=55) (N=84) (N=139) Histological Type ## and Grade ## Oligodendroglioma II 65 (23%) 38 (45%) 21 (15%) 6 (11%) Oligodendroglioma III 44 (16%) 31 (37%) 6 (4%) 7 (13%) Oligoastrocytoma II 41 (15%) 9 (11%) 30 (22%) 2 (4%) Oligoastrocytoma III 33 (12%) 4 (5%) 20 (14%) 9 (16%) Astrocytoma II 30 (11%) 1 (1%) 24 (17%) 5 (9%) Astrocytoma III 65 (23%) 1 (1%) 38 (27%) 26 (47%) Age (Yrs) at Diagnosis ## Mean (SD) 42.6 (13.5) 45.4 (13.2) 38.1 (10.9) 49.9 (15.3) Min, Max 14, 75 17, 75 14, 70 21, 74 Gender Male 155 (56%) 45 (54%) 84 (60%) 26 (47%) Race (self-report, n=274) White 261 (95%) 79 (98%) 131 (95%) 51 (93%) Ethnicity (self-report, n=261) Hispanic/Latino 14 (5%) 5 (6%) 6 (5%) 3 (6%) Treating Country United States* 265 (95%) 81 (96%) 131 (94%) 53 (96%) Year of Diagnosis Before 2005 38 (14%) 10 (12%) 18 (13%) 10 (18%) 2005-2009 88 (32%) 30 (36%) 44 (32%) 14 (25%) 2010-2013 152 (55%) 44 (52%) 77 (55%) 31 (56%) Family History of Cancer (n=190**)# No History 108 (56%) 30 (52%) 64 (65%) 13 (38%) Primary Brain 11 (6%) 2 (3%) 7 (7%) 2 (6%) Other Cancers 72 (38%) 26 (45%) 27 (28%) 19 (56%) Extent of resection (n=268) Open Biopsy 6 (2%) 1 (1%) 4 (3%) 1 (2%) Subtotal Resection 98 (37%) 31 (38%) 45 (34%) 22 (40%) Gross Total Resection 164 (61%) 49 (60%) 83 (63%) 32 (58%) Tumor Location## Frontal Lobe 172 (62%) 68 (81%) 84 (60%) 20 (36%) Parietal Lobe 23 (8%) 5 (6%) 13 (9%) 5 (9%) Temporal Lobe 74 (27%) 9 (11%) 40 (29%) 25 (45%) Other+ 9 (3%) 2 (2%) 2 (1%) 5 (9%) Laterality (n=276) Left 133 (48%) 37 (44%) 69 (50%) 27 (49%) Midline 5 (2%) 2 (2%) 2 (1%) 1 (2%) Right 138 (50%) 45 (54%) 66 (48%) 27 (49%) White vs. Grey Matter (n=144) White Matter 74 (51%) 26 (54%) 37 (51%) 11 (46%) First Presenting Symptom (n=252) Headache 64 (25%) 15 (21%) 39 (30%) 10 (20%) Mental Status Change 22 (9%) 7 (10%) 10 (8%) 5 (10%) Motor/Movement Change 18 (7%) 6 (8%) 7 (5%) 5 (10%) Seizure 135 (54%) 38 (53%) 70 (54%) 27 (53%) Sensory Change 6 (2%) 3 (4%) 1 (1%) 2 (4%) Visual Change 7 (3%) 3 (4%) 2 (2%) 2 (4%) Use of Preoperative Antiseizure Yes 148 (73%) 48 (75%) 73 (72%) 27 (75%) Med. (n=202) Use of Preoperative Corticosteroids Yes 90 (43%) 29 (43%) 46 (44%) 15 (43%) (n=207)

£ Eleven cases with clinical information do not have IDH/Codel status determined * Twelve cases were submitted from Russia (3 IDHmut-codel, 7 IDHmut-non-codel, 2 IDHwt), and three case from Italy (1 IDHmut-non- codel, 2 unknown IDH/codel) ** Cases with response to both questions of family history of any cancer (n=192) and of family history of primary brain cancer (n=197) + One case (IDHwt) was in the cerebellum, three cases were in the occipital lobe (2 IDHmut-codel, 1 IDHmut-non-codel) and five cases were listed as “supratentorial, not otherwise specified” (1 IDHmut-non-codel, 4 IDHwt)

Table 3.1: Clinical characteristics of the sample set. Clinical characteristics are presented in total and by IDH mutant status and 1p/19q co-deletion status within the IDH mutant group. The number of samples with known information is given when complete data are not available. Count, and percentage within group, is given for categorical variables and distributions are compared by Fishers exact test. Mean, SD, and range are given for age at diagnosis and ANOVA was used to compare groups. Signiﬁcance is noted as } 0.10< p <0.05, # p<0.05, ## p<0.01.

46 for a second level cluster of clusters (CoC) analysis. CoC deﬁnes a consensus clustering of the 254 LGG with molecular data from all 4 platforms [90] and resulted in 3 CoC clusters representing distinctive biological themes (Figures 3.1 and A.3). We found a strong correlation between CoC cluster assignment and molecular subtype as deﬁned by IDH mutation and 1p/19q co-deletion status: the large majority of IDHwt LGG are in the CoC cluster with overlapping mRNA R2, microRNA Mi3, DNA methylation M4 and DNA copy number C2 groups.

Another CoC cluster contains almost all IDHmut-codel LGG and includes primarily R3,

M2 and M3, and C3 clusters. The third CoC cluster was highly enriched for IDHmut- non-codel LGG and includes R1, M5, C1 and Mi1 classes.

To determine the relative strength of clinical LGG classiﬁcation schemes in captur- ing the biologic subsets revealed by CoC analysis, we next compared IDH-co-deletion subtype to histology for correlation with CoC cluster assignment. Among 282 samples with both IDH/co-deletion status and CoC cluster assignments, 127/141 (90%) of

IDHmut-non-codel tumors mapped to CoC1; 44/56 (79%) of IDHwt tumors to CoC2; and 82/85 (96%) of IDHmut-codel tumors to CoC3. In contrast, comparison of histology to CoC (293 samples) showed that 62/100 (62%) of astrocytomas mapped to

CoC1; 51/77 (66%) of the oligoastrocytomas also mapped to CoC1; and 72/116 (62%) of the oligodendrogliomas mapped to CoC3. None of the histologies showed predomi- nant mapping to CoC2. Importantly, the concordance between IDH-codel status and

CoC cluster assignment was much greater than that of histology (adjusted Rand indices

47 Age <= 32 Age <= 32 32 < Age < 41 32 < Age < 41 Age >= 41 Age >= 41 Male Male Female Female Grade 2 Grade 2 Grade 3 Grade 3 Astrocytoma Astrocytoma Age <= 32 Oligoastrocytoma Age <= 32 Oligoastrocytoma Age32 < <=Age 32 < 41 Oligodendroglioma 32 < Age < 41 Oligodendroglioma 32Age < >=Age 41 < 41 IDHmut−codel Age >= 41 IDHmut−codel AgeMale >= 41 IDHmut−non−codel Male IDHmut−non−codel MaleFemale IDHwt Female IDHwt FemaleGrade 2 Consensus Cluster 1 Grade 2 Consensus Cluster 1 Grade 23 Consensus Cluster 2 Grade 3 Consensus Cluster 2 GradeAstrocytoma 3 Consensus Cluster 3 Astrocytoma Consensus Cluster 3 AstrocytomaOligoastrocytoma Member Oligoastrocytoma Member OligoastrocytomaOligodendroglioma Non−Member Oligodendroglioma Non−Member OligodendrogliomaIDHmut−codel IDHmut−codel IDHmut−codelnon−codel IDHmut−non−codel IDHmutIDHwt −non−codel Grade IDHwt IDHwtConsensus Cluster 1 Histology Consensus Cluster 1 Molecular Subtypemir_2 Consensus Cluster 12 Consensus Cluster 2 Consensus Cluster Consensus Cluster 23 Consensus Cluster 3 mRNA_1 ConsensusMember Cluster 3 Member mir_2mi2 MemberNon−Member methyl_5 Non−Member mRNA_1R1 Non−Member cn_1 methyl_5M5 mir_4mir_2 cn_1C1 methyl_1 mir_4mi4 mir_3 methyl_1M1 mRNA_1methyl_2 mir_3mi3 mRNA_3 methyl_2M2 mRNA_2 mRNA_3R3 cn_2 methyl_5 mRNA_2 methyl_4R2 cn_2 mir_1C2 methyl_4 mRNA_4M4 mir_1cn_1 methyl_3mi1 mRNA_4 cn_3R4

methyl_3M3 TCGA − P5 A5EX TCGA − HT 7601 TCGA − DU 7007 TCGA − HT 8106 TCGA − HT 8563 TCGA − HT 7686 TCGA − DB 5277 TCGA − E1 5304 TCGA − DU 6404 TCGA − HT 7854 TCGA − FG 7643 TCGA − DU 7292 TCGA − HT 7477 TCGA − DU 7010 TCGA − P5 A5F6 TCGA − DU 7290 TCGA − DU 6406 TCGA − DU 7013 TCGA − HT 7860 TCGA − DB A64O TCGA − HT A617 TCGA − HT A61C TCGA − HW A5KK TCGA − HT 8110 TCGA − FG 6692 TCGA − DU 8165 TCGA − FG 6688 TCGA − DU 7006 TCGA − CS 6186 TCGA − HT 8104 TCGA − CS 6188 TCGA − HT A5RC TCGA − HT A5RA TCGA − FG A4MU TCGA − FG A4MW TCGA − CS 5397 TCGA − CS 5395 TCGA − DU 7012 TCGA − DU 8161 TCGA − DU A5TT TCGA − DU A5TY TCGA − CS 4941 TCGA − DH 5140 TCGA − HT 8011 TCGA − DU 6403 TCGA − DU 6402 TCGA − DU 6405 TCGA − DU 5847 TCGA − DU 8158 TCGA − DU 5852 TCGA − DU 5854 TCGA − DU 6392 TCGA − HT 7882 TCGA − HT 7857 TCGA − HT 7469 TCGA − HT A4DS TCGA − DH A66F TCGA − QH A65V TCGA − FG 7634 TCGA − DU 6400 TCGA − DU 6393 TCGA − DU 7294 TCGA − HT A5R9 TCGA − DB A64L TCGA − FG 7641 TCGA − HT A615 TCGA − DB A64U TCGA − DB A64W TCGA − DB A64V TCGA − DB A64Q TCGA − DB A64P TCGA − DB A64R TCGA − HT 7481 TCGA − HT 7480 TCGA − HT 8105 TCGA − DU 5870 TCGA − HT 7681 TCGA − DU 5874 TCGA − HW 7495 TCGA − DU 7302 TCGA − DH A66B TCGA − DU 8164 TCGA − FG 5964 TCGA − HT 7471 TCGA − HT 7687 TCGA − E1 5319 TCGA − E1 5318 TCGA − E1 5311 TCGA − HW 8322 TCGA − FG 6690 TCGA − DU 7009 TCGA − CS 6668 TCGA − FG A60K TCGA − HT 7468 TCGA − HT A4DV TCGA − HW 7491 TCGA − CS 5396 TCGA − CS 5394 TCGA − CS 5390 TCGA − HT 7616 TCGA − DU 7018 TCGA − DB A4XA TCGA − DB A4XG TCGA − DU 8168 TCGA − DB A4XH TCGA − FG 8186 TCGA − HT 7692 TCGA − HT 7693 TCGA − HT 7677 TCGA − FG 7637 TCGA − HT 7608 TCGA − DH 5141 TCGA − DU 6397 TCGA − HT A619 TCGA − HT 7877 TCGA − QH A65Z TCGA − HT 8012 TCGA − HW A5KJ TCGA − DU 6394 TCGA − P5 A5F0 TCGA − DU 5849 TCGA − HT 7875 TCGA − EZ 7264 TCGA − P5 A5ET TCGA − IK 8125 TCGA − DH 5144 TCGA − DB 5274 TCGA − FG 7638 TCGA − DB 5279 TCGA − HT 7684 TCGA − FG 5965 TCGA − HT 7607 TCGA − HT 7856 TCGA − HT 7467 TCGA − HT 7881 TCGA − FG 8187 TCGA − HT 7695 TCGA − HT 7605 TCGA − HT 8010 TCGA − HT 7874 TCGA − HT 7620 TCGA − HW 7486 TCGA − DU 7300 TCGA − FG 5962 TCGA − HT 8109 TCGA − HW 7487 TCGA − HT 7694 TCGA − HW 8319 TCGA − DU A5TP TCGA − E1 5302 TCGA − HT 7610 TCGA − DB A4X9 TCGA − HT 7472 TCGA − DB 5278 TCGA − DB 5275 TCGA − HT 8564 TCGA − DU 8162 TCGA − HT 7478 TCGA − DU 5872 TCGA − FG 5963 TCGA − HT 7680 TCGA − HT 7691 TCGA − DB A64S TCGA − HT A61A TCGA − CS 6665 TCGA − DU 7304 TCGA − HT 8113 TCGA − HT 7476 TCGA − DU 7309 TCGA − HT 7688 TCGA − HT 8558 TCGA − CS 6669 TCGA − FG 8181 TCGA − HT 8107 TCGA − HT 8019 TCGA − HT 8015 TCGA − P5 A5EY TCGA − FG 8189 TCGA − DU 7014 TCGA − CS 6667 TCGA − DU 8167 TCGA − DU 7298 TCGA − DU 7306 TCGA − DU 7301 TCGA − HT 7855 TCGA − DB 5281 TCGA − FG 8188 TCGA − E1 5305 TCGA − E1 5307 TCGA − HW 8320 TCGA − DU 7008 TCGA − HT 7884 TCGA − DU 7019 TCGA − DB A4XF TCGA − DB A4XD TCGA − FG 8185 TCGA − P5 A5EZ TCGA − DU A5TR TCGA − HT 7604 TCGA − HT 7611 TCGA − E1 5322 TCGA − HW A5KM TCGA − DU 6399 TCGA − HW A5KL TCGA − P5 A5F4 TCGA − DU 6395 TCGA − DU 7015 TCGA − HT 7609 TCGA − DH 5142 TCGA − FG 8191 TCGA − HT A614 TCGA − P5 A5F2 TCGA − FN 7833 TCGA − DB 5280 TCGA − HT 7483 TCGA − E1 5303 TCGA − DU 5871 TCGA − DU 8166 TCGA − HW 7490 TCGA − HT 7676 TCGA − HT 8114 TCGA − HT A5R7 TCGA − HT A5R5 TCGA − QH A65S TCGA − HT 8108 TCGA − FG 8182 TCGA − FG 6689 TCGA − HT A61B TCGA − FG 6691 TCGA − CS 6666 TCGA − HT A618 TCGA − HT A616 TCGA − FG A4MY TCGA − FG A4MX TCGA − HT A5RB TCGA − FG A4MT TCGA − CS 6290 TCGA − CS 5393 TCGA − DB A4XC TCGA − DB A4XB TCGA − DB A4XE TCGA − HT 7690 TCGA − DU A5TS TCGA − DU A5TW TCGA − IK 7675 TCGA − DU 6542 TCGA − P5 A5EV TCGA − HT 7606 TCGA − CS 4943 TCGA − CS 4942 TCGA − CS 4944 TCGA − HT 7689 TCGA − HT 7475 TCGA − HT 7473 TCGA − P5 A5EU TCGA − P5 A5EW TCGA − DU 6408 TCGA − DU 6401 TCGA − P5 A5F1 TCGA − FG A60J TCGA − HW 7489 TCGA − HT 7879 TCGA − FG 7636 TCGA − DB 5273 TCGA − DB 5276 TCGA − DH 5143 TCGA − HT 7873 TCGA − DU 6396 TCGA − DU 5853 TCGA − HW 7493 TCGA − HT 7603 TCGA − HT 7474 TCGA − HT 8018 TCGA − HT 7470 TCGA − HT 7902 TCGA − HT 7880 TCGA − DU 5851 TCGA − DU 6407 TCGA − HT 7602 TCGA − HT 7858 TCGA − HT 7482 TCGA − HT 7485 TCGA − DU 8163 TCGA − CS 4938 TCGA − HT 8111 TCGA − HT 7479 TCGA − HT 8013 TCGA − HW 8321 TCGA − DU 6410 TCGA − DB A64X TCGA − DU A5TU TCGA − DU 5855 TCGA − DU 7299 mir_4cn_3C3 TCGA − P5 A5EX TCGA − HT 7601 TCGA − DU 7007 TCGA − HT 8106 TCGA − HT 8563 TCGA − HT 7686 TCGA − DB 5277 TCGA − E1 5304 TCGA − DU 6404 TCGA − HT 7854 TCGA − FG 7643 TCGA − DU 7292 TCGA − HT 7477 TCGA − DU 7010 TCGA − P5 A5F6 TCGA − DU 7290 TCGA − DU 6406 TCGA − DU 7013 TCGA − HT 7860 TCGA − DB A64O TCGA − HT A617 TCGA − HT A61C TCGA − HW A5KK TCGA − HT 8110 TCGA − FG 6692 TCGA − DU 8165 TCGA − FG 6688 TCGA − DU 7006 TCGA − CS 6186 TCGA − HT 8104 TCGA − CS 6188 TCGA − HT A5RC TCGA − HT A5RA TCGA − FG A4MU TCGA − FG A4MW TCGA − CS 5397 TCGA − CS 5395 TCGA − DU 7012 TCGA − DU 8161 TCGA − DU A5TT TCGA − DU A5TY TCGA − CS 4941 TCGA − DH 5140 TCGA − HT 8011 TCGA − DU 6403 TCGA − DU 6402 TCGA − DU 6405 TCGA − DU 5847 TCGA − DU 8158 TCGA − DU 5852 TCGA − DU 5854 TCGA − DU 6392 TCGA − HT 7882 TCGA − HT 7857 TCGA − HT 7469 TCGA − HT A4DS TCGA − DH A66F TCGA − QH A65V TCGA − FG 7634 TCGA − DU 6400 TCGA − DU 6393 TCGA − DU 7294 TCGA − HT A5R9 TCGA − DB A64L TCGA − FG 7641 TCGA − HT A615 TCGA − DB A64U TCGA − DB A64W TCGA − DB A64V TCGA − DB A64Q TCGA − DB A64P TCGA − DB A64R TCGA − HT 7481 TCGA − HT 7480 TCGA − HT 8105 TCGA − DU 5870 TCGA − HT 7681 TCGA − DU 5874 TCGA − HW 7495 TCGA − DU 7302 TCGA − DH A66B TCGA − DU 8164 TCGA − FG 5964 TCGA − HT 7471 TCGA − HT 7687 TCGA − E1 5319 TCGA − E1 5318 TCGA − E1 5311 TCGA − HW 8322 TCGA − FG 6690 TCGA − DU 7009 TCGA − CS 6668 TCGA − FG A60K TCGA − HT 7468 TCGA − HT A4DV TCGA − HW 7491 TCGA − CS 5396 TCGA − CS 5394 TCGA − CS 5390 TCGA − HT 7616 TCGA − DU 7018 TCGA − DB A4XA TCGA − DB A4XG TCGA − DU 8168 TCGA − DB A4XH TCGA − FG 8186 TCGA − HT 7692 TCGA − HT 7693 TCGA − HT 7677 TCGA − FG 7637 TCGA − HT 7608 TCGA − DH 5141 TCGA − DU 6397 TCGA − HT A619 TCGA − HT 7877 TCGA − QH A65Z TCGA − HT 8012 TCGA − HW A5KJ TCGA − DU 6394 TCGA − P5 A5F0 TCGA − DU 5849 TCGA − HT 7875 TCGA − EZ 7264 TCGA − P5 A5ET TCGA − IK 8125 TCGA − DH 5144 TCGA − DB 5274 TCGA − FG 7638 TCGA − DB 5279 TCGA − HT 7684 TCGA − FG 5965 TCGA − HT 7607 TCGA − HT 7856 TCGA − HT 7467 TCGA − HT 7881 TCGA − FG 8187 TCGA − HT 7695 TCGA − HT 7605 TCGA − HT 8010 TCGA − HT 7874 TCGA − HT 7620 TCGA − HW 7486 TCGA − DU 7300 TCGA − FG 5962 TCGA − HT 8109 TCGA − HW 7487 TCGA − HT 7694 TCGA − HW 8319 TCGA − DU A5TP TCGA − E1 5302 TCGA − HT 7610 TCGA − DB A4X9 TCGA − HT 7472 TCGA − DB 5278 TCGA − DB 5275 TCGA − HT 8564 TCGA − DU 8162 TCGA − HT 7478 TCGA − DU 5872 TCGA − FG 5963 TCGA − HT 7680 TCGA − HT 7691 TCGA − DB A64S TCGA − HT A61A TCGA − CS 6665 TCGA − DU 7304 TCGA − HT 8113 TCGA − HT 7476 TCGA − DU 7309 TCGA − HT 7688 TCGA − HT 8558 TCGA − CS 6669 TCGA − FG 8181 TCGA − HT 8107 TCGA − HT 8019 TCGA − HT 8015 TCGA − P5 A5EY TCGA − FG 8189 TCGA − DU 7014 TCGA − CS 6667 TCGA − DU 8167 TCGA − DU 7298 TCGA − DU 7306 TCGA − DU 7301 TCGA − HT 7855 TCGA − DB 5281 TCGA − FG 8188 TCGA − E1 5305 TCGA − E1 5307 TCGA − HW 8320 TCGA − DU 7008 TCGA − HT 7884 TCGA − DU 7019 TCGA − DB A4XF TCGA − DB A4XD TCGA − FG 8185 TCGA − P5 A5EZ TCGA − DU A5TR TCGA − HT 7604 TCGA − HT 7611 TCGA − E1 5322 TCGA − HW A5KM TCGA − DU 6399 TCGA − HW A5KL TCGA − P5 A5F4 TCGA − DU 6395 TCGA − DU 7015 TCGA − HT 7609 TCGA − DH 5142 TCGA − FG 8191 TCGA − HT A614 TCGA − P5 A5F2 TCGA − FN 7833 TCGA − DB 5280 TCGA − HT 7483 TCGA − E1 5303 TCGA − DU 5871 TCGA − DU 8166 TCGA − HW 7490 TCGA − HT 7676 TCGA − HT 8114 TCGA − HT A5R7 TCGA − HT A5R5 TCGA − QH A65S TCGA − HT 8108 TCGA − FG 8182 TCGA − FG 6689 TCGA − HT A61B TCGA − FG 6691 TCGA − CS 6666 TCGA − HT A618 TCGA − HT A616 TCGA − FG A4MY TCGA − FG A4MX TCGA − HT A5RB TCGA − FG A4MT TCGA − CS 6290 TCGA − CS 5393 TCGA − DB A4XC TCGA − DB A4XB TCGA − DB A4XE TCGA − HT 7690 TCGA − DU A5TS TCGA − DU A5TW TCGA − IK 7675 TCGA − DU 6542 TCGA − P5 A5EV TCGA − HT 7606 TCGA − CS 4943 TCGA − CS 4942 TCGA − CS 4944 TCGA − HT 7689 TCGA − HT 7475 TCGA − HT 7473 TCGA − P5 A5EU TCGA − P5 A5EW TCGA − DU 6408 TCGA − DU 6401 TCGA − P5 A5F1 TCGA − FG A60J TCGA − HW 7489 TCGA − HT 7879 TCGA − FG 7636 TCGA − DB 5273 TCGA − DB 5276 TCGA − DH 5143 TCGA − HT 7873 TCGA − DU 6396 TCGA − DU 5853 TCGA − HW 7493 TCGA − HT 7603 TCGA − HT 7474 TCGA − HT 8018 TCGA − HT 7470 TCGA − HT 7902 TCGA − HT 7880 TCGA − DU 5851 TCGA − DU 6407 TCGA − HT 7602 TCGA − HT 7858 TCGA − HT 7482 TCGA − HT 7485 TCGA − DU 8163 TCGA − CS 4938 TCGA − HT 8111 TCGA − HT 7479 TCGA − HT 8013 TCGA − HW 8321 TCGA − DU 6410 TCGA − DB A64X TCGA − DU A5TU TCGA − DU 5855 TCGA − DU 7299

methyl_1 Figure 3.1: Multiplatform analyses point to biologic subtypes deﬁned by IDH mutation and 1p/19q codeletion status. Cluster of clusters analysis uses the cluster assignments derivedmir_3 from individual molecular platforms to stratify tumors, thereby integrating data from analysis of mRNA (green), miRNA (purple), DNA methylation (blue) and copy number (red). For each sample, membership in a particular cluster is indicated bymethyl_2 a yellow tick; non-membership by a blue tick.

mRNA_3

mRNA_2

cn_2

48 methyl_4

mir_1

mRNA_4

methyl_3

cn_3 TCGA − P5 A5EX TCGA − HT 7601 TCGA − DU 7007 TCGA − HT 8106 TCGA − HT 8563 TCGA − HT 7686 TCGA − DB 5277 TCGA − E1 5304 TCGA − DU 6404 TCGA − HT 7854 TCGA − FG 7643 TCGA − DU 7292 TCGA − HT 7477 TCGA − DU 7010 TCGA − P5 A5F6 TCGA − DU 7290 TCGA − DU 6406 TCGA − DU 7013 TCGA − HT 7860 TCGA − DB A64O TCGA − HT A617 TCGA − HT A61C TCGA − HW A5KK TCGA − HT 8110 TCGA − FG 6692 TCGA − DU 8165 TCGA − FG 6688 TCGA − DU 7006 TCGA − CS 6186 TCGA − HT 8104 TCGA − CS 6188 TCGA − HT A5RC TCGA − HT A5RA TCGA − FG A4MU TCGA − FG A4MW TCGA − CS 5397 TCGA − CS 5395 TCGA − DU 7012 TCGA − DU 8161 TCGA − DU A5TT TCGA − DU A5TY TCGA − CS 4941 TCGA − DH 5140 TCGA − HT 8011 TCGA − DU 6403 TCGA − DU 6402 TCGA − DU 6405 TCGA − DU 5847 TCGA − DU 8158 TCGA − DU 5852 TCGA − DU 5854 TCGA − DU 6392 TCGA − HT 7882 TCGA − HT 7857 TCGA − HT 7469 TCGA − HT A4DS TCGA − DH A66F TCGA − QH A65V TCGA − FG 7634 TCGA − DU 6400 TCGA − DU 6393 TCGA − DU 7294 TCGA − HT A5R9 TCGA − DB A64L TCGA − FG 7641 TCGA − HT A615 TCGA − DB A64U TCGA − DB A64W TCGA − DB A64V TCGA − DB A64Q TCGA − DB A64P TCGA − DB A64R TCGA − HT 7481 TCGA − HT 7480 TCGA − HT 8105 TCGA − DU 5870 TCGA − HT 7681 TCGA − DU 5874 TCGA − HW 7495 TCGA − DU 7302 TCGA − DH A66B TCGA − DU 8164 TCGA − FG 5964 TCGA − HT 7471 TCGA − HT 7687 TCGA − E1 5319 TCGA − E1 5318 TCGA − E1 5311 TCGA − HW 8322 TCGA − FG 6690 TCGA − DU 7009 TCGA − CS 6668 TCGA − FG A60K TCGA − HT 7468 TCGA − HT A4DV TCGA − HW 7491 TCGA − CS 5396 TCGA − CS 5394 TCGA − CS 5390 TCGA − HT 7616 TCGA − DU 7018 TCGA − DB A4XA TCGA − DB A4XG TCGA − DU 8168 TCGA − DB A4XH TCGA − FG 8186 TCGA − HT 7692 TCGA − HT 7693 TCGA − HT 7677 TCGA − FG 7637 TCGA − HT 7608 TCGA − DH 5141 TCGA − DU 6397 TCGA − HT A619 TCGA − HT 7877 TCGA − QH A65Z TCGA − HT 8012 TCGA − HW A5KJ TCGA − DU 6394 TCGA − P5 A5F0 TCGA − DU 5849 TCGA − HT 7875 TCGA − EZ 7264 TCGA − P5 A5ET TCGA − IK 8125 TCGA − DH 5144 TCGA − DB 5274 TCGA − FG 7638 TCGA − DB 5279 TCGA − HT 7684 TCGA − FG 5965 TCGA − HT 7607 TCGA − HT 7856 TCGA − HT 7467 TCGA − HT 7881 TCGA − FG 8187 TCGA − HT 7695 TCGA − HT 7605 TCGA − HT 8010 TCGA − HT 7874 TCGA − HT 7620 TCGA − HW 7486 TCGA − DU 7300 TCGA − FG 5962 TCGA − HT 8109 TCGA − HW 7487 TCGA − HT 7694 TCGA − HW 8319 TCGA − DU A5TP TCGA − E1 5302 TCGA − HT 7610 TCGA − DB A4X9 TCGA − HT 7472 TCGA − DB 5278 TCGA − DB 5275 TCGA − HT 8564 TCGA − DU 8162 TCGA − HT 7478 TCGA − DU 5872 TCGA − FG 5963 TCGA − HT 7680 TCGA − HT 7691 TCGA − DB A64S TCGA − HT A61A TCGA − CS 6665 TCGA − DU 7304 TCGA − HT 8113 TCGA − HT 7476 TCGA − DU 7309 TCGA − HT 7688 TCGA − HT 8558 TCGA − CS 6669 TCGA − FG 8181 TCGA − HT 8107 TCGA − HT 8019 TCGA − HT 8015 TCGA − P5 A5EY TCGA − FG 8189 TCGA − DU 7014 TCGA − CS 6667 TCGA − DU 8167 TCGA − DU 7298 TCGA − DU 7306 TCGA − DU 7301 TCGA − HT 7855 TCGA − DB 5281 TCGA − FG 8188 TCGA − E1 5305 TCGA − E1 5307 TCGA − HW 8320 TCGA − DU 7008 TCGA − HT 7884 TCGA − DU 7019 TCGA − DB A4XF TCGA − DB A4XD TCGA − FG 8185 TCGA − P5 A5EZ TCGA − DU A5TR TCGA − HT 7604 TCGA − HT 7611 TCGA − E1 5322 TCGA − HW A5KM TCGA − DU 6399 TCGA − HW A5KL TCGA − P5 A5F4 TCGA − DU 6395 TCGA − DU 7015 TCGA − HT 7609 TCGA − DH 5142 TCGA − FG 8191 TCGA − HT A614 TCGA − P5 A5F2 TCGA − FN 7833 TCGA − DB 5280 TCGA − HT 7483 TCGA − E1 5303 TCGA − DU 5871 TCGA − DU 8166 TCGA − HW 7490 TCGA − HT 7676 TCGA − HT 8114 TCGA − HT A5R7 TCGA − HT A5R5 TCGA − QH A65S TCGA − HT 8108 TCGA − FG 8182 TCGA − FG 6689 TCGA − HT A61B TCGA − FG 6691 TCGA − CS 6666 TCGA − HT A618 TCGA − HT A616 TCGA − FG A4MY TCGA − FG A4MX TCGA − HT A5RB TCGA − FG A4MT TCGA − CS 6290 TCGA − CS 5393 TCGA − DB A4XC TCGA − DB A4XB TCGA − DB A4XE TCGA − HT 7690 TCGA − DU A5TS TCGA − DU A5TW TCGA − IK 7675 TCGA − DU 6542 TCGA − P5 A5EV TCGA − HT 7606 TCGA − CS 4943 TCGA − CS 4942 TCGA − CS 4944 TCGA − HT 7689 TCGA − HT 7475 TCGA − HT 7473 TCGA − P5 A5EU TCGA − P5 A5EW TCGA − DU 6408 TCGA − DU 6401 TCGA − P5 A5F1 TCGA − FG A60J TCGA − HW 7489 TCGA − HT 7879 TCGA − FG 7636 TCGA − DB 5273 TCGA − DB 5276 TCGA − DH 5143 TCGA − HT 7873 TCGA − DU 6396 TCGA − DU 5853 TCGA − HW 7493 TCGA − HT 7603 TCGA − HT 7474 TCGA − HT 8018 TCGA − HT 7470 TCGA − HT 7902 TCGA − HT 7880 TCGA − DU 5851 TCGA − DU 6407 TCGA − HT 7602 TCGA − HT 7858 TCGA − HT 7482 TCGA − HT 7485 TCGA − DU 8163 TCGA − CS 4938 TCGA − HT 8111 TCGA − HT 7479 TCGA − HT 8013 TCGA − HW 8321 TCGA − DU 6410 TCGA − DB A64X TCGA − DU A5TU TCGA − DU 5855 TCGA − DU 7299 0.79 versus 0.20 Table S2E), suggesting that IDH-codel status provides a simple and accurate reflection of specific LGG biologic subsets. Taken together, integrated analysis of 4 genome-wide platforms resulted in three robust and concordant LGG biologic subtypes whose membership could be largely defined by their IDH and 1p/19q co-deletion status; correlation with histologic classes was poor by comparison.

3.3.3 Mutational landscape of LGG

We generated a consensus mutation set based on three mutation calling algorithms (see

Methods in Supplemental text), yielding 9,885 mutations detected in 289 samples and corresponding to a frequency of 0.66 mutations per Mb in coding regions (median, 29 mutations/sample; range, 0-597). IDHwt LGG exhibited a higher number of mutations per sample (median, 45) than IDHmut-codel LGG (median, 27; p<0.001) or IDHmut- non-codel LGG (median, 28; p<0.001) (Figures S11-13). LGG mutation frequencies were lower than those for GBM; higher than those for medulloblastoma; and intermediate in the spectrum of mutation frequencies reported for TCGA malignancies (see

Figure S11) [70]. Mutations in LGG subsets were dominated by C > T transitions, as shown in Figure S12. One IDHwt sample had a relatively high number of mutations

(597), possibly associated with somatic missense mutations in DNA repair genes POLE and MLH1.

We identiﬁed statistically signiﬁcant DNA copy number alterations and gene mutations

49 for each of the three molecular subtypes (Figures 3.2A, 3.3, S14A, A.1; Tables S3, S4).

Consistent with prior ﬁndings [10] we found CIC mutations in 62% and FUBP1 in 29% of LGG in the IDHmut-codel group, but not in other molecular subtypes, and also observed mutations in PI3 kinase pathway genes, PIK3CA (20%) and PIK3R1 (9%).

Other mutations in this subtype included NOTCH1 (31%) [10, 64, 158], and novel mutations in ZBTB20 (9%), and ARID1A (6%)(Figure 3.2A). Among IDHmut-codel samples, TERT promoter mutations were present in 96%; ATRX mutations were extremely rare in this subset, consistent with prior studies showing mutual exclusivity of these mutations [65]. TERT expression was signiﬁcantly higher in IDHmut-codel than

IDHmut-non-codel LGG, as expected from an activating mutation (Figure 3.2B). The

IDHmut-codel subtype showed focal ampliﬁcation 19p13.3 (Figures 3.2B, S14A), but few other recurring whole arm copy number alterations aside from 1p and 19q loss (Fig- ure S8B). Diﬀerences in mutational frequency and copy number alterations in grades

II and III were modest (Figures 3.2, 3.3A, S21A). Overall, integrated genomic analysis suggests that one route of LGG initiation starts with IDH mutation, followed by 1p/19q co-deletion and TERT activation with deactivation of CIC, FUBP1, and activating mutations/ampliﬁcations in the PI3K pathway. NOTCH1 mutations in IDHmut-codel

LGG were most likely inactivating, since they occurred in sites similar to inactivating mutations in lung, head and neck and cervical cancers (Figure S15), but not at activation sites, as described in T cell acute lymphoblastic leukemia [150] (Figure S15).

PARADIGM-Shift [92] analysis (Figure S16), which evaluates downstream targets to assess pathway status, also suggested NOTCH1 mutations were inactivating in LGG

50 [22]. Prior studies have identiﬁed NOTCH1 mutations in both oligodendroglioma and anaplastic astrocytoma [10, 64, 158]. We identiﬁed them most frequently in IDHmut- codel LGG, yet also rarely in IDHmut-non-codel and IDHwt LGG (Figure 3.2A).

Among 141 IDHmut-non-codel cases, 94% harbored inactivating TP53 mutations and

86% had inactivating alterations of ATRX, including mutations (79%), deletions (3%), gene fusion (2%) or two aberrations (2%). TERT promoter mutations were rare (4%) in this group consistent with the alternative lengthening of telomeres (ALT) associated with ATRX mutations [65]. Concurrent somatic alteration of IDH1, ATRX and TP53 suggests the combined deregulation of metabolism, chromatin organization and apoptosis pathways, respectively, as a mechanism of LGG development in this subtype. The mutually exclusive mutational proﬁles observed between the IDHmut-codel and IDHmut non-codel subtypes likely reﬂect distinct molecular mechanisms of LGG oncogenesis and suggest a near requirement for either 1p/19q co-deletion or TP53 mutation following

IDH mutation. While prior studies have shown enrichment of oligodendrogliomas for

1p/19q codeletion and of astrocytomas for TP53 mutations, with oligoastrocytomas showing weaker associations with these markers, we demonstrate that IDH mutant LGG are either 1p/19q codeleted or TP53 mutated, with minimal overlap or gaps, indicating a strict molecular dichotomy in IDH mutant gliomas [4].

Two signiﬁcantly mutated genes in the IDHmut-non-codel LGG have not been previously reported: the SWI/SNF chromatin remodeler SMARCA4 (6%), implicated in glioma progression [58]; and the translation initiation factor EIF1AX (<1%), for which

51 somatic mutations have been found in uveal melanoma [84] (Figure 3.2A; Table S4).

IDHmut-non-codel LGG showed focal gains of 4q12, a locus harboring the receptor tyrosine kinase (RTK) PDGFRA; 12q14, involving cell cycle regulator CDK4; and 8q24, possibly targeting MYC (Figure S14A), analogous to prior ﬁndings in IDH1-mutant

(MYC) and IDH1-wt (CDK4/PDGFRA) proneural GBM [15]. Histologic grade III tumors in this subset had greater frequencies of chromosome 9p and 19q losses, as well as 10p gains, suggesting these events may occur with progression (Figure 3.3A). Mu- tational proﬁles in this subset did not diﬀer substantially between grades II and III

(Figure S21B).

3.3.4 Signaling networks in LGG

To use a multiplatform approach for classifying LGG that also incorporates their mutational landscape, we performed OncoSign analysis [27] using 64 selected functional genetic and epigenetic events. This approach identiﬁed four dominant tumor subtypes

(OSC1, 2, 3 and 4), which again largely recapitulated those deﬁned by IDH mutation and 1p/19q status (adjusted Rand index, 0.83) (Figure 3.2B; Table S2E). OSC1 showed strong correlation with IDHmut-non-codel and OSC4 contained exclusively ID-

Hwt LGG. The IDHmut-codel group contained both OSC2 and OSC3 LGG, which diﬀered by the presence of mutations in CIC, FUBP1 and NOTCH1, yet were similar in terms of tumor grade and patient outcome. The concordance of IDH mutation and 1p/19q status with two multiplatform genomic data integration approaches (CoC

52 and OncoSign) is striking and contrasts sharply with the much weaker relationship between histology and unsupervised multiplatform classes (Table S2E). Given the high inter-observer variability of histopathologic diagnosis [1, 144], the finding that clinically available markers (IDH, 1p/19q) are capable of classifying LGG in a manner similar to the unsupervised stratification of genome-wide data suggests that IDH and 1p/19q status should be implemented as clinical classifiers of LGG on a routine basis.

3.3.5 IDHwt LGG and GBM

Seven gene mutations were highly associated with IDHwt LGG and ﬁve of these seven were previously reported in the TCGA analysis of GBMs [15]: PTEN (23%), EGFR

(27%), NF1 (20%), TP53 (14%) and PIK3CA (9%). We also found novel mutations in protein tyrosine phosphatase PTPN11 (7%) and phospholipase C gamma-1 (PLCG1;

5%) (Figure 3.2A). Similarly, copy number alterations in IDHwt tumors were distinct from those of IDH-mutant tumors, and instead resembled IDHwt GBMs [15] (Figure

3.3A). In particular, chromosome 7 gain and 10 loss co-occurred in > 50% of tumors

in this subtype (chr7: 56% and chr10: 63%), yet were absent in IDHmut groups. Re-

curring focal ampliﬁcations containing EGFR, MDM4 and CDK4 (38%, 13% and 7%

respectively) and focal deletions targeting CDKN2A and RB1 (63% and 25%) were the

most frequent acquired copy number variants in the IDHwt subtype, and also mirrored

frequencies in IDHwt GBM (Figure 3.3B). These ﬁndings raise the possibility that some

clinically identiﬁed IDHwt LGG represent surgically undersampled GBMs; conversely,

53 !0(! !"#$%&'!()%&

*+,-./ *+,-./%&1*1% *+,-./ 1+75/'%082(% B 234"-%&56!1,- "!05 !"#$ !"#% !"#& !"#' !()*"+,(-#./0010 A 8>7#*B ?>7/)*$P(&-$"%Q R"8>7

=%2(/>,R1*. /+&,#$*.#

8447 0"Q'&*/U-$-S$(2* =&,#$*.S%TS+&,#$*.#

G.-U* !"#$%&%'( )%&*+,&-./0,1$(2*# 3456 7389 :;2:6/<==>?;@ AB;C/<)D=@ ;B6EF6/<7G93@ H?!: ?>@A >J3=!: KLM4: :N2:6F6/<43M4:@ =+0)7+>+(/(3-/.317/3+*(0- N2;:/<=?O>;7@ IGK8 43I> :B6;/<)?)C@

7389 R8>7 3I83 *W2.*##"%Q

2+03*.*,4 5*.1)6./7-0683491 2+03*.*,+)/.-?7/@1 :.317/3+*(-;491 7#$.%+($%R- H?!V$ G; !"'P/7R2&"T"+-$"%Q J&"'%-#$.%+($%R- H?!R,$SQ%QS+%U*& G6 G*Q*/K,#"%Q J&"'%U*QU.%'&"%R- H?!R,$S+%U*& !"'P/?*&*$"%Q 0%R-$"+/),$-$"%Q "+(,.1-=/3/-;491-#.603170 R8>7/*W2.*##"%Q =%2(/Q,R1*./ ?>7/R*$P(&-$"%Q R"8>7/*W2.*##"%Q 8447 #.60317-*<-).603170 8: 86 =: =6 ): )C R": R"6 4: 46 =%=: =%=6 8; 8C =; ); )5 R"; R"C 4; 4C =%=; )6

Figure 3.2: Muational landscape and unbiased clustering. Panel A shows the mutational landscape of somatic alterations in LGG. a) Somatic mutation rates per patient stratified by non-synonymous (blue) and synonymous mutations (green). b) Clinical features associated with patients (see color legend). c) Genes significantly mutated (MutSig2CV q-value < 0.1) in LGG. Mutation types are indicated in specific colors (see color legend). Gray indicates no information available. Panel B illustrates a) Unbiased clustering of tumors based on recurrent copy number alterations, mutations, and gene fusions identifies four main classes (OncoSign Classes or OSCs). White indicates no information available. Classes are largely consistent with the molecular subtypes identified by IDH1/2 mutations and 1p/19q arm losses, and are associated with single data type clusters and tumor histology. b) Combinations of selected events, termed oncogenic signatures, characterize each OSC. A small group of samples showed none of the recurrent events used in this analysis and were therefore considered unclassified. c) TERT promoter mutation and gene overexpression was found to be mutually exclusive with loss of ATRX, consistent with the hypothesis that both alterations have a similar effect on telomere maintenance.

54 the genomic similarity of IDHwt LGG to IDHwt GBM could also indicate that IDHwt

LGG represent an immediate precursor. In either case, IDHwt status in a surgically sampled diﬀuse glioma lacking histologic criteria for GBM indentiﬁes an aggressive tumor, most often with clinical and genetic attributes of GBM; this represents a dramatic improvement over a diagnostic practice based on histology alone.

3.3.6 Genomic rearrangements and fusion transcripts

We investigated 20 samples with high level whole genome sequencing, 50 samples with low-pass whole genome sequencing and 311 samples with sequencing of coding regions only, for the presence of structural chromosomal variants such as translocations and inversions. This analysis uncovered 250 high-conﬁdence chromosomal rearrangements, of which 110 had both ends of the breakpoint within a protein-coding gene (Table

C.3). Additionally, 19 samples showed evidence of extrachromosomal DNA structures known as double minute chromosomes/breakpoint-enriched regions (DM/BER) (Table

C.3, Figure A.2). Of these, 15 occurred in IDHwt samples (27%) (Figure 3.3A), similar in frequency to that in GBM (23%) [15, 121, 161]. Analysis of RNA sequencing data [141, 87] identiﬁed fusion transcripts in 265 LGG tumors (Table S6). Correlat- ing fusion events to structural genomic variants suggested chimeric transcription for

44% of the high-confidence chromosomal rearrangements including two fusions involving EGFR (Figure 3.3B), and 58% of the DM/BER events. Thirty-seven transcript fusion events involved protein kinases, and eleven affected significantly mutated genes.

55 Several genes (EGFR, FGFR3, NOTCH1, ATRX, and CDK4) were aﬀected by fusions in multiple samples (Figure 3.3C). Activating RTK fusions of EGFR and FGFR3, previously found in GBMs [15, 128, 101], were restricted to IDHwt LGG at frequencies comparable to GBM (7% and 3%, respectively; Figures 3.3C, S18, S19) [15]. A chimeric and novel FGFR3-ELAVL3 transcript involved the same breakpoint as previously reported FGFR3-TACC3 fusions and was highly expressed, suggesting similar eﬀects on

FGFR3 function. Three samples had fusions between EGFR and intergenic or intronic regions on chromosome 7 that are predicted to remove the EGFR autophosphorylation domain and are likely oncogenic (Figure 3.3C) [26]. Gene fusions involving RTKs were predominantly a feature of IDHwt LGG, as only two IDHmut-non-codel tumors harbored RTK fusions (PDGFRA and MET), and none were identiﬁed in IDHmut-codel samples.

3.3.7 LGG protein expression

Reverse-phase protein lysate array (RPPA) analysis resulted in protein expression pro-

ﬁles that demonstrated a striking segregation of IDHwt from IDHmut LGG, and a consistent pattern of RTK pathway activation within IDHwt tumors, including EGFR phosphorylation, providing additional support that IDHwt LGG are biologically similar to GBM (Figures 3.3D, S20). The overexpression of HER2 in this subset represents a potential therapeutic target. Among the IDHmut-LGG, higher expression of hematopoietic markers Syk, E-cadherin and Annexin1 was noted in the non-codel

56 group, whereas higher phosphorylated HER3 (PY1289) was identiﬁed in the IDHmut- codel group. HER3 overexpression may be associated with resistance to PI3K inhibitors

[42].

3.3.8 From signatures to pathways

To gain insights into signaling pathways, we examined mutations, focal copy number alterations, structural variants and gene fusions aﬀecting RTKs (EGFR, PDGFRA, MET,

FGFR), PI3 kinase subunits, MAP kinases, NF1 and BRAF, components of the p53 and RB1 pathways (MDM2, MDM4, MDM1, CDKN2A/B and CDKN2C), and ATRX.

Alterations across these loci were remarkably similar in frequency between IDHwt LGG and IDHwt GBMs, but not other LGG subtypes (Figure 3.3B, 3.4A). Forty-three percent of IDHwt LGG and 53% of IDHwt GBM tumors harbored an alteration in EGFR, with EGFR ampliﬁcation being the most frequent in both (Figure S21). Homozygous

CDKN2A/B deletions occurred in 45% of IDHwt LGG, similar in range to IDHwt GBM

(55%). This contrasts with IDHmut-non-codel LGG, in which large single copy deletions of chromosome 9p were common, yet CDKN2A/B was homozygously deleted in only 4% (Figure A.1). IDHmut tumors lacked IDHwt-GBM-associated cancer pathway aberrations and instead had their own characteristic cancer pathway alterations that included TP53 and ATRX (IDHmut-non-codel) and TERT, NOTCH1, CIC and FUBP1

(IDHmut-codel) (Figure 3.4A). Grade II IDHwt LGG were uncommon (13 cases), yet diﬀered in their mutational proﬁle from grade III (Figure 3.3A, S21C, D). Grade II

57 / / / * * * $ $ $ * * * G G G G G G F F F R R R ' ' ' S S S D D D \ \ \ W W W D D D Q Q Q V V V H H H X X X W W W P P P V V V E E E H H H U U U

J J J L L L V V V W W W L L L F F F B B B W W W K K K U U U H H H V V V 7 7 7 K K K & & & R R R * * * O O O G G G $ $ $ H H H E E E G G G U U U

D D D L L L Q Q Q O O O R R R Z Z Z H H H U U U J J J U U U D D D G G G H H H J J J O O O L L L R R R P P P D D D

/ / / * * * * * *

F F F R R R S S S \ \ \ Q Q Q X X X P P P E E E / / / H H H / / / * * * * * * $ * $ $ * * U U U $ $ $ G * * * G G G G G G G G F F F G G G R F F F R R ' ' ' J J J R R R S S S ' ' ' D D D S S S \ \ \ W D D D W W L L L D \ \ \ D D W W W V V Q V V D D D Q Q V V H Q Q Q V V V H H X X X W H H H W W W W W X X X P V P P V V W W W P P P V V V L L L E E E F F F E E E H / / / H H * * H H H * U U U $ $ U U U $ * * *

G G G J

J J B B B G G G F F F J J J L L L R R R V ' ' ' V V L L L S S S W W W V V V W D D W W D L \ \ \ L L W W W W W W F F F D D K K D K L L L F F F Q Q V V Q V H H H X X X B B B W W W U U U P P V V P B B B V W W W K K K W W W H H H E E E K K K U U U H H H H H H U U U U U V V U V H H H V 7 V V 7 7

K V V V 7 7 7 K K & J J & & J K K K K K K R & & & R R L L L * * * V V V R R R O O O * * * W W W G R R $ G G R $ $ L L O O O L F F G G G F H $ $ $ H H E H H H G E E O O G G O B B B E E E U G G G

U U G G G

W W W D D D U U U K K

K D D D L L L U U U Q H H Q Q H H H L L L H Q Q Q V V 7 7 V 7 O O O K K K R & & G G R R & G O O O R R R R R R Z * * * Z Z O O O Z Z Z G G G $ $ $ H H H H H H H H H U H H H U U E E E G G G U U U U U U J

J J D D D V V V J J J U U U L L L D Q Q Q D D U U U W W D D D W G G G O O O R R G G G R H L L H H L Z Z Z H H H P P P J J J H H H J J J O U U U O O L L L O O O R / / / R R J J J L L L * * * D D R R R D P U U P P U $ $ $ * * * D D G G G D P P P G G G F F F D G G G D D R R R W W W ' ' ' S S S D D D H H H D D D \ \ \ W W W

H H ! H

D D D J J Q Q Q / J V V V

/ / H H H X X X ! ! O O / / / O * W W W * * ! P P P L L V V V L R R R * * * E E E * * * P P P H H H * * * U U U IDHmut- codel IDHmut- non-codel IDHwt GBM IDHwt

D D D

J J J F ! F F L L L

V V V

R F F F W W W / / R R / L L L ! !

F F F !""## R R R

* * !""# !""# "23# S * ! S S B B B S S S $#!" \ * * $%&)*(+,-,+.-/01# * \ \ $%&)*(+.-/01# $%&'(# $%&'(# IDHmut- IDHmut- W W % Samples With W K K K \ \ \ IDHmut- codel IDHmut- non-codel IDHwt GBM IDHwt

IDHwt

U U U Q H H H Q Q codel non-codel Alteration ! F F V V V 7 7 7 F $!!" A Q Q Q B 100 X K K K X X 1p del 1p & & & R R R R R R ! ! X X X * * * P P P S S O O O S ! G G G $ $ $ P P P #!" \ \ H H H \ 50 E E E Chr 1 E G G G E E $#!" U U U

Q Q IDHmut- codel IDHmut- non-codel IDHwt GBM IDHwt E E E Q D D D H H H L L L !" X X Q Q Q X H H H

0 U ! U U P P $!!" P O O O U U U R R R 7 amp 7 J ! ! J J Z Z Z ! E E E %&'()*" J J J H H H L +,-"$!"012" +,-"$3"012" +,-".3"012" L L +,-"4"563" V +,-"$./"012" U U U H H V V #!" Chr 7 H 50 L L L $#!" ./0"'6"345" J J J W V V V U U W W U IDHmut- codel IDHmut- non-codel IDHwt GBM IDHwt U U U L L L W W W D D D F J J F F J ! G G G L L L ./0"7"896" 0 !" F F F $!!" L L H H H L V V V 9p del 9p ! ! J J J B B B W W ! W O O O L L L L L L B B B W %&'()*"

./0"16"345" R R R F F W W F #!" +,-"$!"012" +,-"$3"012" +,-".3"012" Chr 9 +,-"4"563" 50 P P P +,-"$./"012" K $#!" K K W W IDHmut-codel! W IDHmut- codel IDHmut- non-codel IDHwt GBM IDHwt D D D K K K U B B B U U ./0"'!"345" IDHmut-non-codel ! H

!" H H U U U W W

W 0 / / / $!!" K K H H H K * * * V 10 del 10 V V ! IDHwt! * * * U U U K V V V

./0"'12"345" K K H H ! ! H

 %&'()*" GBM IDHwt! K K K R +,-"$!"012" +,-"$3"012" +,-".3"012" +,-"4"563" ! #!"

Chr 10 R R 50 F F F V V +,-"$./"012" V $#!" R R R R R R O K K K ()*+,-" O O S S S G G G O O O R R \ \ \ R

IDHmut- codel IDHmut- non-codel IDHwt GBM IDHwt 0 !" G G G H H H O O $!!" Q Q Q O

100 del 19q G G !" #!" $!" %!"G &!" '!!" '#!" H H H X X X G G G P P P H H H G G G %&'()*" G G E E E G #!" H +,-"$!"012" +,-"$3"012" +,-".3"012" +,-"4"563"

Chr 19 H H 50 +,-"$./"012" H H H H H H V U U U V V H H H $#!" J J J W V V V W W V V V !" L L L

0 L V V V L L W W W W W W P W W W P P L L L L L L L L L F F F P P P P P P Grade 2 3 2 3 2 3 $!!" DM/HSR D D D %&'()*" B B B +,-"$!"012" +,-"$3"012" +,-".3"012" +,-"4"563" D D D W W W +,-"$./"012" D D D W W W K K K W W W #!"

50 H U U U H H W W W H H H H H H H H H V V V

DM K K K !" R R 0 R O O O G G G H H H

G G G ;90/12&*2;*60,-)(6*8%<=*6-(/%./*0)<(9012&*%&*>(&(* %&'()*" +,-"$!"012" +,-"$3"012" +,-".3"012" +,-"4"563" H H H +,-"$./"012" V V V W W W

L L L !"#$%&'()* +,-)%./012&* 3()(12&* !#* 456%2&* *782*29*,29(*0:(99012&6* P P P D D D W W W H H H C D

Figure 3.3: IDHwt LGG resemble IDHwt GBMs. Panel A shows the frequency of large-scale copy number alterations in specific LGG molecular subtypes, which have been divided by histologic grade. The UCSC Cancer Genomics Browser [45] (https://genome- cancer.ucsc.edu) was used to visualize GISTIC thresholded copy number calls across the indicated chromosomes. Each horizontal line shows the copy number for an individual sample colored red (amplification), blue (deletion), or white (normal) at each genomic position. Percentages for the indicated event (far right, written vertically) are shown in the bar charts on the right. IDHwt LGG have similar frequencies of gains and losses as IDHwt GBMs and are distinct from IDHmut LGG. Panel B shows the frequencies of mutational events commonly found in IDHwt GBM in the indicated LGG molecular subtypes, IDHmut-codel (N = 85), IDHmut-non-codel (N = 141), and IDHwt (N = 56). Differences in mutational frequency by tumor grade are shown in Figure S21. Panel C shows schematic representations of the activating receptor tyrosine kinase fusions involving EGFR and FGFR3 that were identified in IDHwt LGG tumors. The UCSC Cancer Genomics Browser view of exon-level expression for the fusion partners is displayed for each fusion-positive patient. The exon-level expression tracks are lined up with the fusion breakpoints in each case. Panel D shows the relative total and phos- phoprotein expression levels from reverse phase protein array (RPPA) analysis. From a cohort of 189 proteins, the 40 proteins shown here were found to be differentially expressed (BH-adjusted p-value < 1e-5) among the three molecular subtypes. Expres- sion levels of most of the receptor tyrosine kinase pathway proteins (highlighted) except phosphoHER3 are significantly elevated in IDHwt LGG compared to other subtypes. All of the samples with EGFR amplification or fusion, and most of the samples with EGFR mutation showed elevated EGFR expression levels.

58 IDHwt LGG were highly enriched with tumors from the small, discrete M1 methylation cluster (Figure 3.1, S1A), which display a low frequency of mutations and copy number alterations, and do not have TERT promoter mutations, potentially indicating they are distinct pathologic entities. TERT promoter mutations were present in 64% of the

IDHwt LGG; when M1 LGG are removed, TERT promoter mutations were present in

80% of those remaining, consistent with frequencies in primary GBM [65].

3.3.9 Clinical characteristics and outcomes of molecular subtypes

Patients with IDHwt LGG were older than those with IDHmut LGG and had a more frequent family history of cancer (Table 3.1). Anatomic locations also diﬀered, with

IDHmut LGG arising in the frontal lobes more often than IDHwt. Among 289 cases with clinical follow-up, 77/250 (31%) experienced recurrence, and 60/289 (21%) were deceased at the time of analysis. Consistent with molecular ﬁndings that showed similarity of IDHwt LGG to GBM [50], survival analysis showed that patients with IDHwt

LGG had substantially shorter overall survival than IDHmut LGG (age adjusted haz- ard ratio 7.4, 95% CI (4.0, 13.8)) and had a prognosis similar to IDHwt GBMs (Figure

3.4B, Table S2D). In comparison, LGG in the IDHmut-codel and non-codel groups were associated with relatively prolonged median survivals of 8.0 years and 6.3 years, respectively. Molecular classification of LGG as IDHwt, IDHmut-non-codel and IDHmut-codel stratifies patient outcomes in multi-predictor models after adjusting for age and extent of resection (Tables S2B-D). Grade, but not histological diagnosis, remains a signifi-

59 cant predictor of outcome in multivariable models with IDH/codel status and provides additional prognostic value among the molecular subsets (Table S2; Figure S22). To- gether the results point to three robust LGG subtypes, each with prototypic molecular alterations and distinctive clinical behaviors (Figure 3.4C).

3.4 Discussion

This integrative analysis of 293 diﬀuse LGG employed a comprehensive, multiplatform approach to delineate biological foundations and establish potential therapeutic targets.

The findings are striking in their clarity and consistency: 1) unsupervised classification from diverse genome-wide analytic platforms provides strong biologic justification for subtyping LGG based on IDH and 1p/19q status as an improvement over conventional histologic classification; 2) IDH mutant LGG are either 1p/19q codeleted or TP53 mutant, with very few exceptions; and 3) the majority of IDHwt-LGG showed remarkable genomic and clinical similarity to primary GBM. Our results have several immediate clinical ramifications. Current management of LGG patients is based on histologic classification schemes that suffer from significant interobserver variability [31, 143, 1].

Nevertheless, critical decisions regarding therapy depend on these distinctions [14, 120].

Our ﬁndings distilled the six possible histology and grade combinations into three robust and non-overlapping molecular subtypes, laying the groundwork for a reproducible and clinically relevant classiﬁcation that incorporates molecular data into the patho-

60 A

B C

Figure 3.4: Integrative analysis and clinical outcome. Panel A demonstrates pathway alterations of LGG as a schematic that summarizes genomic alterations-mutations and focal amplifications and homozygous deletions-in LGG across molecular subtypes. Canonical RTK/PI3K/MAPK, RB and p53 regulatory pathways are frequently altered, as is telomere maintenance. However, alterations occur in different pathway components depending on the LGG subtype. Panel B depicts the Kaplan-Meier estimates of overall survival for the LGG samples classified by IDH mutation and 1p/19q co-deletion as well as GBM samples (from previously published TCGA data [15]) classified by IDH mutational status. Age-adjusted analysis can be found in Supplemental Table 2, while further division by histologic grade is present in Figure S22. Overall survival times and IDH mutation status of GBMs in this analysis were as previously described [15]. Panel C summarizes major findings and conclusions, where consensus clustering yields three robust groups based on IDH and 1p/19q co-deletion status with stereotypic and subtype-specific molecular alterations and distinct clinical presentation.

61 logic diagnosis, as planned for the upcoming revision of the WHO classiﬁcation of brain tumors [79]. The primary subtype discriminator for LGG was the mutational status of IDH1/IDH2, with all molecular platforms demonstrating a sharp contrast between

IDHwt and IDHmut tumors.

Whereas IDHwt tumors showed an enrichment of astrocytomas and IDHmut-codel

LGG were predominantly oligodendrogliomas, the largest molecular subclass of LGG

(IDHmut-non-codel) contained all histologic types, highlighting the non-specificity in- herent to pure histologic classification. By comparison, two unsupervised methods using multiplatform analysis to classify LGG (CoC and OncoSign) yielded strong correlations with IDH/codel status (adjusted Rand index 0.79 and 0.83, respectively), suggesting that such molecular classification captures biological classes more accurately than histology. Additionally, while oligodendrogliomas were typically 1p/19q co-deleted and astrocytomas generally were not, oligoastrocytomas were distributed among the three molecular subtypes with no molecular feature distinguishing them. Thus, while previous WHO classification systems have recognized LGG with mixed histology (oligoastrocytoma), molecular classification indicates that IDH mutant LGG are either 1p/19q co-deleted or TP53 mutant in a nearly mutually exclusive fashion, and does not support the existence of this entity biologically (Figure 3.4C) [152, 118]. Not only will the use of molecular signatures for LGG lead to a practice-altering, biologically-based diagnosis with greatly improved interobserver concordance; a phase-out of the diagnosis oligoastrocytoma and the confusion related to its clinical management is a major advance

62 stemming from our results.

Another substantial finding was that IDHwt tumors were molecularly and clinically distinct from IDHmutant subtypes, with most showing a striking resemblance to primary GBM on all analytic platforms. The findings suggest that IDHwt LGG may represent immediate precursors of IDHwt GBM that have similar clinical properties; alternatively, such tumors could represent actual GBMs that were incompletely sampled during surgery, precluding definitive histologic classification. Sampling errors represent a significant challenge in surgical neuropathology, regardless of IDH status, class or grade, since a morphologic diagnosis is limited to the findings under the microscope.

Thus, molecular classiﬁcation based on IDH and 1p/19q status represent an improvement in diagnostic practice, since it enables the identiﬁcation of a clinically aggressive form of LGG (IDHwt) in the absence of morphological criteria for GBM.

Consistent with earlier work [57, 20, 151], analysis of clinical outcomes demonstrated that IDHmut-noncodel LGG performed somewhat less favorably than IDHmut-codel

LGG with both subtypes exhibiting substantially longer overall survival than their ID-

Hwt counterparts. Patients with IDHmut-codel LGG had a median survival of 8.0 years, compared to the IDHwt subgroup, which had median a survival of 1.1 years (Fig- ure 3.4B). Compared to outcome prediction based on histologic classiﬁcation (Figure

S22), stratiﬁcation of clinical risk based on IDH and 1p/19q status is more robust. In addition to reducing inter-observer variability and predicting risk, molecular classiﬁca- tion can provide quality control for histopathologic diagnosis. As an example, the small,

63 discrete DNA methylation cluster M1 in our cohort showed a low frequency of mutations and copy number alterations and contained occasional tumors with BRAF alterations.

While not entirely specific, these alterations are more characteristic of grade I circumscribed tumors, such as pilocytic astrocytoma and ganglioglioma, and its finding would lead to a consideration of alternative diagnoses. Given that histopathologic features of diffuse versus circumscribed gliomas can overlap, while clinical management differs greatly, molecular signatures offer the potential to resolve diagnostically challenging cases. Further analysis of our cohorts survival data as it matures will be required to designate precise metrics by which WHO grade may combine with molecular signatures to optimize risk stratification for LGG.

The variable clinical course of LGG patients, coupled with the three sharply contrast- ing molecular profiles, suggest that distinct therapeutic strategies may be required for effective disease control. Molecular inclusion criteria and stratification in clinical trial design will be necessary for clear interpretation of outcomes from specific treatments.

The prevalence of IDH mutations in LGG invites targeting of either the mutant enzymes themselves or the consequences of downstream metabolic and epigenomic derangement, speciﬁcally G-CIMP, and such eﬀorts are already underway [115, 142]. ATRX, CIC, and

FUBP1 mutations have only recently been implicated in cancer biology; yet their speci-

ﬁcity and prevalence in IDHmut LGG support central roles in oncogenesis and argue for thorough characterization of associated signaling networks to facilitate therapeutic development. The remarkable genetic and clinical similarities between IDHwt-LGG and

64 primary GBM support the potential inclusion of IDHwt LGG within the broad spectrum of GBM-related clinical investigation and treatment protocols. The extent to which ID-

Hwt LGG resemble GBMs in their response to recently developed therapies has yet to be determined. Finally, our integrative analysis has shown that all LGG subtypes rely to some extent on core signaling networks previously implicated in GBM pathogenesis.

Many agents targeting these pathways are currently available and in clinical trials and may prove eﬀective against LGG subtypes.

65 Chapter 4

Mutation Independent Activation of

The Notch Pathway is Associated With

Lapatinib Resistance in Her2+ Breast

Cancer Cell Lines

Cancer can not only progress through genomic alterations, but can also progress through non-genomic resistance to treatment. This chapter describes a mutation independent mechanism of resistance to a commonly used HER2 tyrosine kinase inhibitor, lapatinib, in HER2+ breast cancer cell lines. These ﬁndings show that the Notch pathway is associated with this resistance and suggest that Notch inhibitors may be beneﬁcial in treating patients with lapatinib resistance.

66 This chapter contains text from the paper ”Mutation Independent Activation of the

Notch Pathway is Associated with lapatinib Resistance in Her2+ Breast Cancer Cell

Lines” by Mia Griﬀord, Melanie Plastini, Sol Katzman, Soﬁe R. Salama and David

Haussler in preparation (to be submitted to PLOS One in 2014). I performed the majority of this work with the help of an undergraduate whom I trained, Melanie Plastini, who helped with the cell culture experiments and performed some of the quantitative

RT-PCR and Western blot experiments (sections 4.2.1, 4.2.2 and 4.2.3). Sol Katzman performed the mapping of RNA-sequencing, running RSEM and running DE-Seq (ﬁgure

B.5 and section 4.2.5), and Amie Radenbaugh ran the mutation calling with the tool she developed, RADIA (section 4.2.8), while I analyzed and presented the results.

4.1 Introduction

Breast cancer patients are divided into subtypes based on molecular aberrations and activation of key proteins [90]. These subtype classiﬁcations help guide treatment.

About 20% of patients fall into the HER2+ subtype, which have an activation of the receptor tyrosine kinase human epidermal growth factor 2 (HER2, also known as

ERBB2) [106]. These patients are commonly treated with a monoclonal antibody for

HER2, trastuzumab, or a HER2 tyrosine kinase inhibitor, lapatinib, in combination with chemotherapy and sometimes other non-chemotherapy drugs [108]. Despite these treatments, HER2+ patients have overall poor survival when compared with other subtypes

67 [62, 102]. Often poor patient survival is associated with drug resistance. There have been many proposed mechanisms of resistance to targeted therapies like trastuzumab and lapatinib, including activation of alternate survival signaling pathways, such as insulin growth factor (IGF) or phosphatidylinositide 3-kinase (PI3K) [56, 80, 52, 12], but a complete explanation of resistance remains elusive. An understanding of the mechanisms underlying resistance to targeted therapies is critical for increasing their eﬀectiveness and developing strategies to treat resistant tumors.

Many patients will respond to re-treatment of a drug after displaying resistance and being taken oﬀ the drug for a period of time [66, 156, 23]. These observations suggest that resistance is reversible in these patients and is not associated with resistance- conferring mutations. Indeed, a reversible, epigenetic mechanism of resistance to a targeted therapy was observed in non-small cell lung cancer (NSCLC) cell lines, which had activated receptor tyrosine kinase epidermal growth factor receptor (EGFR) [126].

EGFR (also known as HER1) is in the same family of receptor tyrosine kinase proteins as

HER2 [116] and can be targeted by the EGFR inhibitor erlotinib, which, like lapatinib, is a tyrosine kinase inhibitor. These NSCLC resistant cell lines had active insulin-like growth factor 1 receptor (IGF-1R) signaling, overexpression of the histone demethylase

KDM5A, diﬀerences in bulk chromatin marks and displayed hypersensitivity to histone deacetylase inhibitors. We wanted to determine if a similar, non-mutational, mechanism of resistance occurs in HER2+ breast cancer cell lines, and further to identify speciﬁc signaling pathways involved in this type of resistance.

68 We found that breast cancer cell lines do indeed acquire lapatinib resistance with features similar to the NSCLC cell line resistance to Erlotinib described in [126]. We could not identify expressed mutations in genes known to be involved in drug resistance, suggesting an epigenetic mechanism for resistance. Transcriptome analysis revealed that the Notch pathway is involved in this mutation-independent mechanism of resistance, similar to what has been shown in other models of resistance

[160, 117, 25, 77, 83, 100, 97, 113]. Analysis of downstream targets of Notch suggests that up-regulation of Notch signaling may lead to changes in chromatin structure through activation of the insulin-like growth factor 1 receptor (IGF1R) and lysine (K)- speciﬁc demethylase (KDM) family proteins.

4.2 Methods

4.2.1 Cell culture and generation of lapatinib resistant cell lines

HCC1419 and BT474 (ATCC, CRL-2326: 1/31/11, HTB-20: 4/7/11) were grown in

RPMI containing 1% penicillin streptomycin (PenStrep) and 10% Gibco certiﬁed fetal bovine serum (FBS) (Invitrogen) (RPMI-complete) or RPMI-complete conditioned with mouse embryonic ﬁbroblasts (MEF). MEF-conditioned media was generated by incubating mitomycin-C inactivated MEFs (55,000 cells/cm2) with RPMI-complete for 24 hours before collecting. MEFs were initially plated in DMEM supplemented with 10% FBS and PenStrep (Invitrogen) and were used for a total of 6 days before discarding. MEF-

69 conditioned RPMI was stored at 4 ◦C until the ﬁnal day of collection and then pooled,

ﬁltered and stored at −80 ◦C. To generate lapatinib resistant lines, cells were plated in

MEF-conditioned RPMI-complete. 24 hours after plating, the media was replaced with

MEF-conditioned RPMI-complete supplemented with 1 µM lapatinib. The media was

changed every 3 days. On day 10, drug tolerant persisters (DTPs) were isolated. Two

independent resistant lines for HCC1419 (drug tolerant expanded persisters, DTEPs)

were grown and isolated at passage numbers 7 and 12 (line 1), and 6 (line 2) for RNA

sequencing. For experiments testing sensitivity to tricostatin A (TSA), BT474 cells

were grown in 6-well dishes containing 3 wells each of RPMI-complete, RPMI-complete

supplemented with 75 nM Trichostatin A (TSA), RPMI-complete supplemented with 1

µM lapatinib or RPMI-complete supplemented with both 75 nM TSA and 1 µM lapatinib. Pictures were taken at 9 days or 72 days. Cells were counted 3 times for each picture and averaged, the average of 7 pictures was used in ﬁgure 2d.

4.2.2 Quantitative RT-PCR in drug tolerant persisters

RNA from about 500,000 cells each of untreated HCC1419 cells, HCC1419 cells treated with 1 µM lapatinib for 24 hours and HCC1419 cells treated with 1 µM lapatinib for

9 days (DTPs) was isolated using Trizol following the manufacturers guidlines. Quan- titative real-time PCR was performed using the QuantiTect SYBR Green RTPCR Kit

(Qiagen), using 20 ng of total RNA per reaction. All reactions were set up in triplicate.

Cycling was performed on a Rotor-Gene 6000 (Qiagen) according to the QuantiTect

70 SYBR Green RTPCR Kit instructions. Primers used are: IGF1R forward: ccattct- catgccttggtct, IGF1R reverse: tgcaagttctggttgtcgag, IGFBP3 forward: gcttctgctggt- gtgtggat, IGFBP3 reverse: ggcgtctacttgctctgcat, KDM5A forward: tccagcctttctacc- caatg, KDM5A reverse: cgtaattgctgccactctga, KDM5B forward: cccacctctccatgatgttc,

KDM5B reverse: tagtccctggctgcttgttc, KDM5C forward: cttgtctgcctttcccacat, KDM5C reverse: agcccgaaccttcagcttat, GAPDH forward: tgaggccggtgctgagtatg, GAPDH reverse: tggttcacacccatcacacaaac. Ct values were calculated with Corbett 6000 software and the relative enrichment of GAPDH-normalized gene expression levels in lapatinib treated cells was determined as fold change over the gene expression levels in untreated cells.

4.2.3 Immunoblotting

One and a half to two million cells each of untreated HCC1419 cells, HCC1419 cells treated with 1 µM lapatinib for 24 hours and HCC1419 cells treated with 1 µM lapa-

tinib for 9 days (DTPs) were suspended in NuPAGE LDS sample buﬀer (Invitrogen) at

a concentration of 10,000,000 cells per 1 ml buﬀer. Histones were extracted by incubat-

ing cells in 0.4 N H2SO4 for 30 min, incubating the supernatant in 132 µl TCA for 30

min, washing the histone pellet with 500 µl acetone and repeating TCA incubation and wash (all steps were performed on ice or in cold room). Histone extractions were loaded on NuPAGE (Invitrogen) 4-12% protein gels for SDS-PAGE. Transfer of proteins onto

PVDF membrane was performed as described in the NuPAGE manual. Immunoblotting

71 was performed using antibodies against H3 (Abcam ab1791, 1:2,500), H3K4me3 (Milli- pore 07-473, 1:500), H3K27me3 (Millipore 07-449, 1:3,000) and H3ac (Millipore 06-599,

1:1,000). Blots were incubated with SuperSignal West Dura Extended Duration Sub- strate (Thermo Scientiﬁc) according to the manufacturers instructions and visualized on a Biorad Chemidoc MP system.

4.2.4 RNA sequencing library preparation

RNA sequencing was performed on 2 biological replicates each of untreated parental

HCC1419 cells (passage # 31), 2 DTP replicates and 3 DTEP replicates (line 1 passage

#s 7 and 12 and line 2 passage # 6). RNA was isolated using Trizol following the manufacturers guidelines. RNA was treated with RQ1 DNAseI (Promega) for 1 hour at 37 ◦C and total RNA was cleaned up using the RNAeasy Mini kit (Qiagen). For each sample, the non-ribosomal fraction of 5 µg of total RNA was isolated using a Ribo-Zero rRNA removal Kit (Epicentre) following the manufacturers protocol (Lit. #309-6/2011). For the non-ribosomal fraction of RNA, 200 ng RNA was fragmented by incubating 30 minutes at 98 ◦C in RNA storage buffer (Ambion) and then double stranded (ds) cDNA was synthesized as described previously [104] using dUTP in the second strand synthesis and USER digest before amplification to retain strand specificity. Clean-up steps were performed using RNA Clean & Concentrator or DNA Clean & Concentrator kits (Zymo

Research). Double stranded cDNA was used for library preparation following the Low

Throughput guidelines of the TruSeq DNA Sample Preparation kit (Illumina), with the

72 following additions. Size selections were performed before and after cDNA ampliﬁca- tion on an E-gel Safe Imager (Invitrogen) using 2% E-gel SizeSelect gels (Invitrogen).

The cDNA fraction of 300-500 bp in size (including adapters) was isolated and puriﬁed.

For adapter ligations, 1 µl instead of 2.5 µl of DNA Adapter Index was used. Indexed libraries were pooled and sequenced on the Illumina HiSEQ platform.

4.2.5 RNA sequencing data mapping and analysis

The input Illumina fastq ﬁles consisted of paired end reads with each end containing 100 bp. These were reduced to 80x80 bp by trimming 20 bp from the 3’ end of each read to improve mapping of fragments shorter than 100 nt. Bowtie2 [68] was used to map the fragments against the set of elements from the repeatMasker [131] human library. After

ﬁltering out such fragments, bowtie2 was used to map the remaining fragments against a version of the human transcriptome derived from the UCSC Known Genes track [53] of the GRCh37hg19 human assembly by using the target generation script supplied with RSEM [73]. Only mappings with both read ends aligned were kept. Potential

PCR duplicates (mappings of more than one fragment with identical positions for both read ends) were removed with the samtools rmdup function [74], keeping only one of any potential duplicates. The ﬁnal set of mapped paired end reads for a sample were converted to transcript and gene expression estimates using RSEM [73]. Then, DESeq [3] was used to identify diﬀerentially expressed genes between sample types (table C.5). For input to DESeq [3] all genes with non-zero counts in any sample were considered.

73 4.2.6 Gene expression clustering and pathway enrichment analysis

ConsensusClusterPlus [153] was used to run hierarchical clustering using Pearson distance on RSEM [73] expression values of genes found to be significantly differentially expressed by DESeq [3] (Benjamini-Hochberg adjusted p-value less than 0.1) (fig. B.4).

Expression values were logged (base 2), and each row (gene) was mean centered and scaled by the standard deviation. Gene set enrichment analysis (GSEA) [138] was run on RSEM expression values to ﬁnd all pathways diﬀerentially expressed between untreated cells and DTEPs from the Kyoto Encyclopedia of Genes and Genomes (KEGG)

[60] or the National Cancer Institute Pathway Interaction Database (PID) [122]. All default parameters were used, except to assess signiﬁcance genes were permuted 1000 times and the log base 2 fold change was used as the metric for ranking genes. Values displayed in ﬁgure 3 are logged (base 2), and each row (gene) is mean centered and scaled by the standard deviation.

4.2.7 Accession codes

The RNA-Seq data has been deposited in NCBIs Gene Expression Omnibus [37] and is accessible through GEO Series accession number GSE62074

(http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE62074).

74 4.2.8 Mutation calling with RADIA

Single Nucleotide Variants (SNVs) were identiﬁed by RADIA (RNA and DNA Inte- grated Analysis), a method that typically combines patient matched normal and tumor

DNA whole exome sequencing (DNA-WES) with tumor RNA sequencing (RNA-Seq) for somatic mutation detection [110]. In this case, RADIA was executed on RNA-Seq data alone. Several ﬁlters were used to eliminate false positives while maintaining true positive calls. All SNVs that overlap with common SNPs found in at least 1% of the samples in dbSNP were removed. Each SNV had at least 10 total reads, a minimum of 4 reads and at least 10% of the reads supported the variant. All standard ﬁlters

(e.g. strand bias, positional bias, base quality and mapping quality) were applied. In addition, any SNV that overlapped with pseudogenes or retrogenes were ﬁltered out.

Lastly, a BLAT [63] ﬁlter was applied to guarantee that each read that supported a variant did not align better to another location in the genome.

4.3 Results

4.3.1 Prolonged incubation in lapatinib robustly generates resistant

cell lines

In order to better understand the mechanisms behind lapatinib resistance, lapatinib resistant cell lines were generated through long-term treatment with lapatinib. Two

75 HER2+ cell lines, HCC1419 and BT474 (ﬁg. B.1), were treated with lapatinib continuously. By day 10, untreated cells became completely conﬂuent, while the vast majority of lapatinib treated cells had died. The remaining cells, termed drug tolerant persisters

(DTPs) as in Sharma, et al. [126], were alive but not proliferating. After about 40 to 70 days, these cells began to proliferate and could be expanded through multiple passages, forming lapatinib resistant cell lines termed drug tolerant expanded persisters (DTEPs)

[126] (ﬁg. 4.1, B.2).

4.3.2 Breast cancer DTPs share features with non-small cell lung can-

cer mutation-independent resistant cells

The breast cancer cell line DTPs we generated are similar to non-small cell lung cancer

(NSCLC) DTPs, which are characterized by a reversible, chromatin-mediated, drug- tolerant state [126]. Like NSCLC DTPs, breast cancer cell line DTPs have increased

IGF-1R and IGFBP3 expression, which has been shown to activate IGF-1R [5, 39] (ﬁg.

4.2a). IGF-1R activation has been shown to be necessary for activation of the histone demethylase KDM5A, which also recruits histone deacetylases [93], and is activated in NSCLC DTPs [126]. We measured the RNA expression levels of many KDM5A paralogs, including KDM5A, KDM5B and KDM5C, which share a common domain structure and function [13] (we excluded KDM5D, because it is not expressed in these cells), and found that all were over-expressed in breast cancer DTPs (ﬁg. 4.2b).

76 Untreated Parental DTPs DTEPs Population Drug Tolerant Persisters Drug Tolerant Expanded Persisters (transient quiescent state) (inherited drug tolerant state)

Day 0 1 4 7 10 ... ~40-70

Plate Replace Replace Replace cells media+drug media+drug media+drug every 3 days Replace media, Replace Replace adding drug media+drug media+drug every 3 days

Figure 4.1: Lapatinib resistant cell lines were generated by long-term lapatinib treatment. Timeline for generation of lapatinib resistant cells. Lapatinib sensitive HCC1419 or BT474 cells were treated with 1 µM lapatinib for 9 days to generate drug tolerant persisters (DTPs, middle), after 40-70 days the cells begin to proliferate and form drug tolerant expanded persisters (DTEPs, right), untreated cells plated at the same initial density proliferate rapidly and become conﬂuent by day 10 (Parental population, left). Micrographs show representative images of the indicated populations of HCC1419 cells.

77 Also shown in NSCLC DTPs were changes in chromatin marks, consistent with increased amounts of closed chromatin [126]. Breast cancer DTPs do indeed have slight changes in chromatin marks as well, including increased H3K27 tri-methylation, a closed chromatin mark, and decreased H3K4 tri-methylation and total H3 acetylation, both open chromatin marks (ﬁg. B.3).

Because KDM5A, KDM5B and KDM5C are over-expressed in DTPs and H3 acetylation is decreased in DTPs, changes in chromatin organization mediated by histone acetylation state may be critical for lapatinib resistance in this setting. We hypothesized that a histone deacetylase inhibitor may aﬀect DTP sensitivity to lapatinib, as was shown in

NSCLC DTPs. We treated resistant cells (DTPs) with the histone deacetylase inhibitor,

Trichostatin A (TSA). As was expected, TSA alone did not aﬀect cell growth, but the combination of TSA and lapatinib prevented DTEP formation (ﬁg. 4.2c-d). From these results we can infer that histone deacetylation, and the resulting changes in chromatin, is necessary for resistance.

4.3.3 RNA-Sequencing reveals signiﬁcant changes in gene expression

We then wanted to determine which genes and pathways were involved in drug resistance. We performed RNA sequencing (RNA-seq) on untreated HCC1419 cells,

HCC1419 DTPs, and HCC1419 DTEPs (2 replicates of each, plus an additional time point of one DTEP line - see methods). In order to get gene expression estimates from

78 A BT474 HCC1419 C 20

10 Sample Sample BT+L 10 HCC+L BT+L24 HCC+L24 BT-L HCC-L Untreated 75nM TSA Relative Expression Relative Relative Expression Relative 9 days 9 days 5 5 Relative Expression Relative Expression Relative

0 0

GAPDH IGF1R IGFBP3 GAPDH IGF1R IGFBP3 GAPDH IGF-1R IGFBP3 GAPDH IGF-1RPrimer IGFBP3 B Primer 4

3 2µM Lapatinib 2µM Lapatinib + 75nM TSA 72 days 72 days 2 Sample Sample BT+L HCC+L BT+L24 2 HCC+L24 BT-L HCC-L D Relative Expression Relative Relative Expression Relative

1 Relative Expression Relative Expression Relative 1

0 0

GAPDH KDM5A KDM5B KDM5C GAPDH KDM5A KDM5B KDM5C GAPDH KDM5APrimerKDM5B KDM5C GAPDH KDM5APrimerKDM5B KDM5C DTPs (+lapatinib for 9 days) Parental Population (+lapatinib for 24 hours) Parental Population (-lapatinib)

Figure 4.2: Breast cancer DTPs share features with non-small cell lung cancer DTPs. A,B: Relative expression measured by quantitative reverse-transcriptase PCR of IGF-1R and IGFBP3 (A) or KDM5A, KDM5B and KDM5C (B) normalized to GAPDH in DTPs (green), parental population treated with lapatinib for 24 hours (light purple) and untreated parental population (dark purple) shows activation of IGF-1R and over-expression of KDM5A paralogs in resistant HCC1419 and BT474 cells. C: Representative micrographs of BT474 cells untreated or treated with 75 nM Trichostatin A (TSA) and/or 2 µM lapatinib shows that TSA prevents formation of DTEPs when combined with lapatinib, but TSA alone does not aﬀect cell growth. D: Quantiﬁcation of cell counts from 7 micrographs as in (C) of cells treated with 2 µM lapatinib or 2 µM lapatinib and 75 nM TSA for 72 days. Error bars represent standard deviation.

79 the sequencing data, we used RSEM - RNA-Seq by Expectation Maximization [73]. We then performed differential gene expression analysis using DESeq [3], and found 6,943 significantly differentially expressed genes (fig. B.4). The DESeq results revealed that the replicates were more similar to one another than they were to other samples (fig.

B.5). Surprisingly, DTP samples were more similar to untreated samples than they were to DTEP samples, suggesting that signiﬁcant gene expression changes were necessary for proliferation to occur in resistant cells. In addition, the DESeq results conﬁrmed that IGF-1R pathway genes are over-expressed in resistant cells (Table B.1). We also looked at the KDM family of genes and found that while KDM5A paralogs are tran- siently over-expressed in DTPs, this does not persist in DTEPs. Other KDM genes

(KDM4B and KDM6B) are over-expressed in DTEPs, suggesting a prolonged eﬀect on histone modiﬁcation.

4.3.4 Resistance is mutation independent

We next wanted to explore whether genetic mutations were underlying these changes in gene expression. Because we had performed RNA-Seq, we were able to determine if there were any expressed mutations in genes known to be involved in drug resistance.

We ran a mutation calling algorithm named RADIA [110] to detect expressed, somatic mutations from the RNA-Seq data. Table C.4 lists all expressed mutations in any

DTEP replicate, but unaﬀected in either untreated replicate. We then searched this list for mutation calls in genes that could be involved in resistance. We searched for

80 over 75 genes known to be involved in drug resistance (Table B.2) [52, 12, 47]. We also searched the list for any RAS, RAF, IGF, IGFBP, PI3K, or MAPK genes (not including interacting proteins). Only one resistance-related gene, IGFBP5, contained a mutation call, and this only in a single DTEP replicate, and not in either untreated replicate. This gene is lowly expressed in most samples and, upon further investigation, was found to have some evidence of mutation in all 7 samples (including both untreated and both DTP replicates). The position of this mutation call is also toward the end of the 3’ UTR, which is a highly repetitive region prone to sequencing and mutation calling errors, and there are no other cancer-related mutations at this location found in the Catalogue of Somatic Mutations In Cancer (COSMIC) [41]. These ﬁndings suggest that this mutation call is likely to be an artifact.

Because there were no expressed mutations likely to cause lapatinib resistance, and because DTEP formation was prevented with a histone deacetylase inhibitor and breast cancer DTPs have high IGF-1R, IGFBP3 and KDM5A, KDM5B and KDM5C expression similar to NSCLC DTPs, we conclude that it is likely that the lapatinib resistance is mutation independent and instead due to epigenetic changes as observed in erlotinib resistant NSCLC cell lines [126].

81 4.3.5 The Notch pathway is over-expressed in resistant cells

To identify pathways over-expressed in resistant cells, Gene Set Enrichment Analysis

(GSEA) [138] was performed. We found all pathways that were signiﬁcantly over- expressed in DTEPs when compared with untreated cells from the Kyoto Encyclopedia of Genes and Genomes (KEGG) [60] or the National Cancer Institute Pathway Interac- tion Database (PID) [122]. Both the KEGG and PID Notch pathways were found to be signiﬁcantly over-expressed in DTEPs compared to untreated cells. Notch1 and Notch3 are over-expressed in DTEPs, as well as several genes downstream of Notch, including

HES1, MYC, FOS and JUN [99, 51]. Other signiﬁcantly overexpressed pathways are

Hedgehog, PDGFRA, IL2 and PI3K. Several genes that are in these pathways have previously been shown to interact with Notch and are involved in several other pathways, including JUN, FOS, MYC, EP300 and PIK3CA. All enriched pathways with a FDR q-value below 0.1 are shown in ﬁgure 4.3, and genes involved in 3 or more pathways are highlighted. An alternate method, involving running functional annotation on DESeq

[3] signiﬁcantly diﬀerentially expressed gene lists, using the Database for Annotation,

Visualization and Integrated Discovery (DAVID) [55, 54], found similar results (data not shown).

In order to determine how closely connected the diﬀerentially expressed genes within these pathways were, we used Cytoscape [29, 119, 125, 132] to visualize the pathway connections between all “core” genes in pathways with an FDR q-value below 0.1 (ﬁg.

82 Pathway (FDR q-value) 1;DE 26 rgtlrn xaddpritr ie# asg #6. passage passage persisters #1 #2 line expanded line persisters persisters tolerant expanded expanded drug tolerant tolerant drug L1P7: drug highlighted L2P6: DTEP genes, L1P12: stan- DTEP DTEP all pathways. the #12; #7; for by more passage membership scaled or #1 pathway and 3 line show centered in mean bars occur was side (gene) genes Black be values row to Expression each Heatmap deviation. found and pathways B: cells. dard 2), 10 untreated over-expressed. the (base to in are logged compared genes 0.1. were that DTEPs all pathway below of in q-value the values over-expressed FDR in expression signiﬁcantly a genes [73] with RSEM of [138] normalized % GSEA of the by shows cells chart untreated Bar to cells. compared DTEPs untreated in to compared DTEPs 4.3: Figure

Pathway (FDR q-value) A KEGG ALDOSTERONE (0.063) KEGG HEDGEHOG (0.015) KEGG DIABETES (0.032) Pathways signiﬁcantly enriched inDTEPs enriched signiﬁcantly Pathways KEGG NOTCH (0.035) KEGG TASTE (0.064) PID PDGFRA (0.015) PID REG GR (0.070) PID IL2/PI3K (0.088) PID IL2/PI3K (0.088) PID NOTCH (0.087) PID NOTCH (0.087) PID IL2 (0.020) PID IL2 (0.020) oc,P3 n O/U ahasaeoe-xrse in over-expressed are pathways FOS/JUN and PI3K Notch,

KEGG_NOTCH_SIGNALING_PATHWAY significantlyPathways enriched in DTEPs KEGG_HEDGEHOG_SIGNALING_PATHWAYPID_NOTCH_PATHWAYPID_IL2_1PATHWAY PID_REG_GR_PATHWAYPID_PDGFRAPATHWAYPID_IL2_PI3KPATHWAY KEGG_ALDOSTERONE_REGULATED_SODIUM_REABSORPTIONKEGG_TYPE_II_DIABETES_MELLITUSKEGG_TASTE_TRANSDUCTION % of over-expressed genes inpathway genes over-expressed of % 0 % of over-expressed genes in pathway % of over-expressed

par_mg01 10 20

par_mg03 30 40 EP300 50 MYC

dtep_mgA6 B 83 PIK3CA SOS1 IRS2 IRS1 FOS JUN :Ptwy on ob over-expressed be to found Pathways A: Notch (KEGG) -1.5 −1.5 ColorKey

Notch (PID) Color Key

KEGG_NOTCH_SIGNALING_PATHWAYValue Hedgehog (KEGG) Value 0 PID_NOTCH_PATHWAYdtep_mg07 0

IL2 (PID) 1 KEGG_HEDGEHOG_SIGNALING_PATHWAY1 IL2/PI3K (PID) PID_IL2_1PATHWAY PDGFRA (PID) PID_IL2_PI3KPATHWAY REG GR (PID) par_mg01 PID_PDGFRAPATHWAY Diabetes (KEGG)dtep_mg08 PID_REG_GR_PATHWAY Aldosterone (KEGG) KEGG_TYPE_II_DIABETES_MELLITUSpar_mg03

KEGG_ALDOSTERONE_REGULATED_SODIUM_REABSORPTIONTaste (KEGG) KCNB1 TAS2R13 GRM4 GNB1 TAS2R19 EP300 NCOR2 JAG2 NOTCH3 NOTCH1 JAG1 CREBBP NUMBL CTBP2 HES1 LFNG RFNG MAML1 DTX2 MYC GATA3 NEURL SSPO FURIN PRKX BMP8B WNT9A ZIC2 LRP2 SUFU BMP8A GLI3 WNT7B CSNK1D PIK3CA SOS1 JAK3 GAB2 BCL2 FOS JUN STAT5B IRS2 IRS1 SOCS2 SOCS3 PRKCE RELA RAPGEF1 SRF PRKACA NR4A1 EGR1 NFATC1 FKBP5 IRF1 SMARCC2 PIK3CD PIK3R2 CACNA1B KCNJ11 PRKCD SCNN1G ATP1B4 PRKCA NEDD4L

KEGG_HEDGEHOG_SIGNALING_PATHWAYKEGG_NOTCH_SIGNALING_PATHWAYKEGG_TASTE_TRANSDUCTIONPID_NOTCH_PATHWAY KEGG_TYPE_II_DIABETES_MELLITUSPID_REG_GR_PATHWAYPID_PDGFRAPATHWAYPID_IL2_PI3KPATHWAYPID_IL2_1PATHWAY KEGG_ALDOSTERONE_REGULATED_SODIUM_REABSORPTION KEGG_TASTE_TRANSDUCTION dtep_mgA6 Untreatedpar_mg01 1

dtep_mg07 Untreatedpar_mg03 2

DTEPdtep_mgA6 L1 P7 dtep_mg08 TAS2R19 GNB1 GRM4 TAS2R13 KCNB1 NEDD4L PRKCA ATP1B4 SCNN1G PRKCD KCNJ11 CACNA1B PIK3R2 PIK3CD SMARCC2 IRF1 FKBP5 NFATC1 EGR1 NR4A1 PRKACA SRF RAPGEF1 RELA PRKCE SOCS3 SOCS2 IRS1 IRS2 STAT5B JUN FOS BCL2 GAB2 JAK3 SOS1 PIK3CA CSNK1D WNT7B GLI3 BMP8A SUFU LRP2 ZIC2 WNT9A BMP8B PRKX FURIN SSPO NEURL GATA3 MYC DTX2 MAML1 RFNG LFNG HES1 CTBP2 NUMBL CREBBP JAG1 NOTCH1 NOTCH3 JAG2 NCOR2 EP300 DTEP dtep_mg07L1 P12

DTEPdtep_mg08 L2 P6 WNT7B CSNK1D PIK3CA SOS1 JAK3 GAB2 BCL2 FOS JUN STAT5B IRS2 IRS1 SOCS2 SOCS3 PRKCE RELA RAPGEF1 SRF PRKACA NR4A1 EGR1 NFATC1 FKBP5 IRF1 SMARCC2 PIK3CD PIK3R2 CACNA1B KCNJ11 PRKCD SCNN1G ATP1B4 PRKCA NEDD4L KCNB1 TAS2R13 GRM4 GNB1 TAS2R19 EP300 NCOR2 JAG2 NOTCH3 NOTCH1 JAG1 CREBBP NUMBL CTBP2 HES1 LFNG RFNG MAML1 DTX2 MYC GATA3 NEURL SSPO FURIN PRKX BMP8B WNT9A ZIC2 LRP2 SUFU BMP8A GLI3

par_mg01

par_mg03

dtep_mgA6

dtep_mg07

dtep_mg08 EP300 NCOR2 JAG2 NOTCH3 NOTCH1 JAG1 CREBBP NUMBL CTBP2 HES1 LFNG RFNG MAML1 DTX2 MYC GATA3NEURL SSPO FURIN PRKX BMP8B WNT9A ZIC2 LRP2 SUFU BMP8A GLI3 WNT7B CSNK1D PIK3CA SOS1 JAK3 GAB2 BCL2 FOS JUN STAT5BIRS2 IRS1 SOCS2 SOCS3 PRKCE RELA RAPGEF1 SRF PRKACA NR4A1 EGR1 NFATC1FKBP5 IRF1 SMARCC2 PIK3CD PIK3R2 CACNA1B KCNJ11 PRKCD SCNN1G ATP1B4 PRKCA NEDD4L KCNB1 TAS2R13 GRM4 GNB1 TAS2R19 B.6, supplementary data C.6). “Core” genes are those that contribute to the GSEA [138] enrichment score, and are therefore over-expressed in DTEPs. Gene nodes were colored based on the DESeq [3] assigned log fold changes and all colored genes have a DE-Seq [3]

Benjamini-Hochberg adjusted p-value less than 0.1. Connections between genes were determined using a superimposed pathway, which incorporates pathway information from the National Cancer Institute Pathway Interaction Database, Biocarta, Reactome and KEGG [60, 122, 91, 32]. This analysis revealed that these pathways are highly interconnected, with only 5 genes left unconnected to the rest of the pathway.

Interestingly, of the 5 ligands for Notch, Jag1 and Jag2 are over-expressed in resistant cells, while the other Notch ligands DLL1, DLL3 and DLL4 are not. This is consistent with previous reports, which show that Jag1 over-expression is associated with tumor recurrence in breast cancer patients [113, 30]. Other groups have found Jag1 and Jag2 to be initially lower in Her2+ tumors when compared with other subtypes, and we also found Jag1 and Jag2 to be lower in Her2+ tumors using patient tumor RNA-Seq data from the The Cancer Genome Atlas and the UCSC Cancer Genome Browser [90, 45, 162]

(ﬁg. B.7). Indeed the majority of genes in the Notch pathway that are over-expressed in

DTEPs are expressed at lower levels in Her2+ tumors than other tumors in the TCGA breast cancer cohort (ﬁg. B.7).

84 4.4 Discussion

Because resistance to targeted therapies in Her2+ breast cancer patients is common, we set out to better understand the mechanisms behind resistance to a Her2 tyrosine kinase inhibitor, lapatinib. Our experiments suggest that the mechanism behind lapatinib resistance in Her2+ breast cancer cell lines is mutation-independent and involves changes in chromatin, similar to Erlotinib resistance in non-small cell lung cancer cell lines

[126]. RNA-sequencing revealed many changes in gene expression between untreated and lapatinib resistant cells (DTEPs), but the most signiﬁcant pathways that were over-expressed in resistant cells were Notch and other closely related pathways. The

Notch pathway may be initially activated through over-expression of the Notch ligands

Jag1 and Jag2 in DTPs.

Notch is an integral part of nearly all the over-expressed pathways in resistant cells and

Notch has previously been shown to be involved in drug resistance to several diﬀerent breast cancer drugs (including tamoxiﬁn, aromatase inhibitors, VEGF receptor tyrosine kinase inhibitors, trastuzumab and lapatinib)[160, 117, 25, 77, 83, 100, 97, 113]. Thus,

Notch activity may be key to development of lapatinib resistance in this setting. Because

Notch1 transcriptionally activates IGF-1R [38], which subsequently activates the KDM family of proteins [126, 72, 33, 34], activation of the Notch pathway may be what leads to the changes in chromatin seen in resistant cells (ﬁg. 4.4). Therefore, treating HER2+ tumors with Notch inhibitors alongside HER2-targeted agents may be of therapeutic

85 beneﬁt.

Gamma secretase inhibitors (GSIs) have been shown to inhibit Notch signaling and have been shown to prevent drug resistance in other settings [96]. There are also several clinical trials underway combining GSIs with other targeted therapies in breast cancer and other cancers. These results describe a mutation-independent mechanism of resistance to lapatinib involving Notch signaling, suggesting that combining GSIs with lapatinib could prevent this type of resistance and thus improve and/or extend the eﬀectiveness of lapatinib treatments.

86 JAG1/2

NOTCH1/3

IGF-1R IRS1/2

SOS1 PIK3CA

AP1

FOS JUN NFKB1 RELA HES1 MYC

Proliferation KDM6B

K27 K27 K27 K4 K4 me3 me3 me3 ac me3 me3 ac

Figure 4.4: Notch activation may lead to altered chromatin state. Pathway diagram displaying how the Notch pathway may aﬀect drug resistance through proliferation and chromatin modiﬁcation. Top is a signal sending cell displaying JAG1/2 and bottom is a signal receiving cell displaying NOTCH. Green nodes are genes over-expressed in DTEPs compared to untreated cells with a DE-Seq [3] Benjamini-Hochberg adjusted p-value less than 0.1. Solid lines represent transcriptional regulation, dashed lines represent post-transcriptional regulation. Ac: histone 3 acetylation; k4me3: histone 3 lysine 4 tri-methylation; k27me3: histone 3 lysine 27 tri-methylation.

87 Chapter 5

Conclusions

High-throughput sequencing technology has allowed for great advances in cancer treatment. Much has been learned about the mechanisms of oncogenesis and new targeted therapies are more successful and have less side eﬀects than traditional chemotherapy and radiation. Though, there are still many problems with cancer treatment and there is still much to learn. Many times cancer treatment is unsuccessful due to late diagnosis, patients not being given the optimum treatment for their particular cancer or drug resistance. These three problems are addressed in this thesis. Each chapter demonstrates another way high-throughput sequencing technology was used in order to make advances in these areas.

In chapter 2, a method for identifying and characterizing double minute chromosomes

(DMs) using whole exome sequencing data is described. I used this method to estimate

88 the prevalance and features of DMs in The Cancer Genome Atlas (TCGA) cohorts of both glioblastoma multiforme (GBM) patients and lower grade glioma patients (LGG).

I also used whole genome sequencing to validate predictions made from whole exome sequencing in GBM patients and to reconstruct a DM in an LGG patient. These circular amplicons containing oncogenes explain the mechanism behind cancer progression in about a quarter of all GBM patients and about a quarter of IDH wild-type lower grade glioma (LGG) patients. And, because double minute chromosomes may be detectable in the blood, they present a new opportunity for early diagnosis and monitoring disease progression without needing surgery.

The method I used to detect double minute chromosomes can now be applied to many other types of cancer. In fact, another graduate student in the Haussler lab is working on just that. There are both exome and whole genome sequencing data for large cohorts of patients (generally between 200 to 1,000 patients) for many different cancer types from The Cancer Genome Atlas, and Arjun Rao has been searching for double minute chromosomes in several of these cancers. Initial searches are done using whole exome data, for which there are many more samples, and potential double minute chromosomes are then verified and reconstructed using whole genome data. Once many cancer types have been analyzed, it will give us a greater understanding of how widespread this phenomenon is. We will be able to determine specifically what types of cancers these double minute chromosomes occur in, if and how the features of these DMs vary across cancer types and what genes are involved. This method, if applied more universally, has

89 the potential to aﬀect diagnosis and monitoring of disease progression in many diﬀerent types of cancer.

In chapter 3, the comprehensive characterization of LGG patients is described. From which a new method of classiﬁcation of samples was presented, which more accurately categorizes patients into treatment groups and identiﬁes a group of patients with poor prognosis, who are similar to GBMs and should receive more aggressive treatment. I performed several analyses that support this new method of subtyping patients, including clustering of clusters from various data types, comparing copy number alterations between LGG and GBM patients, comparing alterations in tyrosine kinases between

LGG and GBM patients, analyzing genomic rearrangements and analyzing diﬀerences in survival between LGG and GBM patients.

Although we have learned a lot about lower grade glioma and how similar it can be to glioblastoma, there is still more to learn. One major question remaining is, if the

IDH wild-type LGG patients are so similar to GBM, then why are they being clas- siﬁed as LGGs? What is diﬀerent between IDH wild-type LGGs and IDH wild-type

GBMs? Likewise, we can still ask what are the differences between IDH mutant LGGs and IDH mutant GBMs (which tend to have much better prognosis)? A new TCGA working group was developed to address these questions and many others. They are specifically working on comparing LGG and GBM tumors in more depth than has ever been previously achievable. They are utilizing many different data types, including exome sequencing, whole genome sequencing, RNA-sequencing, copy number, microRNA

90 sequencing, methylation and reverse phase protein arrays, to analyze cohorts of LGG and GBM patients together. These analyses will lead to a greater understanding of not only the diﬀerences between LGG and GBM, but also what causes LGG patients to progress to GBM. Understanding the mechanisms behind progression may allow for new treatments to be used in order to prevent progression to GBM. This could save many lives, because patients can live long, healthy lives with LGG, but once it progresses to

GBM, the average survival time is only fourteen months.

In chapter 4, I explain how I developed lapatinib resistant cell lines from HER2+ breast cancer lines, and used these resistant lines to gain a better understanding of the mechanism of resistance to the targeted therapy, lapatinib. I describe a mutation-independent mechanism of drug resistance to lapatinib, which may also be applicable in other settings

(diﬀerent drugs or types of cancer). This mechanism of resistance involves activation of the Notch pathway and therefore may be treatable with gamma secretase inhibitors that are already in clinical trials.

Though, more work will need to be done to determine if combining lapatinib with gamma secretase inhibitors will be successful. We are currently testing if a gamma secretase inhibitor, LY-411575, prevents growth of DTEPs when combined with lapatinib, or if it makes cells more sensitive to lapatinib. We also hope that these ﬁndings will encourage other doctors and scientists to continue to study this mechanism of resistance, possibly beginning new clinical trials combining lapatinib with gamma secretase inhibitors. In addition, the RNA sequencing data and the lapatinib resistant cell lines are publicly

91 available and can be used by other groups who are studying similar mechanisms of resistance. One group in particular, at UCSF, is doing drug screenings with lapatinib resistant DTPs to ﬁnd drug combinations that would prevent growth of resistant cells.

They may be able to utilize the DTEP lines I’ve generated in their screens, which would model a diﬀerent type of resistance than the cells they are currently using. If this mechanism of resistance is indeed treatable with Notch inhibitors, it could help save many patients’ lives. Also because a similar mechanism of resistance was already shown to occur in non-small cell lung cancer, this mechanism of resistance should be studied in other cancer types with other targeted therapies. This could allow for the discovery of new treatment combinations, potentially aﬀecting a much broader range of patients.

The projects described in this thesis will help to advance current cancer treatment in several ways. We have gained new insight into the mechanism of oncogenesis in a subset of glioblastoma patients which may allow for earlier diagnosis and monitoring of disease progression without the need for surgery. A new method of characterizing lower grade glioma patients, if adopted by the World Health Organization, will make classifying patients more accurate and will lead to more appropriate treatment, especially for the group of LGG patients with poor prognosis. Finally, describing the method of resistance to the targeted therapy, lapatinib, could lead to new, more eﬀective treatment options for patients with resistance. Each of these projects describes one small step toward advancing treatment for cancer patients and ultimately improving patient outcome. In

92 addition, there are many ways scientists are continuing to build upon this work and utilize these ﬁndings to help larger cohorts of cancer patients.

93 Appendix A

Chapter 3 Supplementary Figures

94 7&*$EUDLQORZHUJUDGHJOLRPD /** FRS\QXPEHUJLVWLFBWKUHVKROGHGHVWLPDWH $GG'DWDVHWV 

/**FRS\QXPEHU JLVWLFBWKUHVKROGHG

IDHmut-non-codel IDHmut-codel IDH wt IDH

CDKN2A/B

Figure A.1: The UCSC Cancer Genomics Browser view of GISTIC thresholded copy number calls for the region of chromosome 9 containing CDKN2A/B. Samples are sorted based on molecular subtype. About 30% of IDHmut- non-codel samples have deletions in CDKN2A/B, similar to the 29% of samples with chr 9p deletions shown in figure 3.3A, while only 4% passed the cutoff (-2) used for gene-level calls made in figure 3.3B.

95 Figure A.2: Reconstruction of TCGA-CS-5395 circular amplicon Panel A shows a tumor browser view of the region of chromosome 1 containing the DM/BER. Com- mon features of DMs/BERs are labeled in purple. The bottom track shows protein coding genes within the region. The track above shows relative copy number of the tumor DNA compared with normal, showing distinct blocks of elevated copy number with similar total copy number between blocks. The top two tracks show intra and interchromosomal rearrangement breakpoints. All rearrangements shown are supported by at least 100 discordant reads. The type of rearrangement is indicated by the color of the line: duplication (red), deletion (blue) and inversion (yellow and green). Panel B shows a diagram of the ampliﬁed segments and structural variants identiﬁed on chromosome 1. Walking through this diagram results in a circular solution, suggesting the double minute chromosome diagrammed in panel C, where segments inverted relative to their orientation in the reference genome are colored blue. The letters inside the circle correspond to the segments in B. The numbers inside the circle indicate the number of sequencing reads supporting each breakpoint. Oncogenes are colored red.

96 ● 0.98 0.97

0.96 ● ● ● 0.95

● cophenetic correlation coefficient 0.94 0.93

●

2 3 4 5 6 7

● 0.90

● 0.85 ● ● 0.80 0.75 average silhouette width average ● 0.70

● 0.65

2 3 4 5 6 7

Figure A.3: Metrics for cluster of clusters model selection. From top-left: consensus cumulative distribution function (CDF), consensus matrix for k=3, cophenetic correlation coeﬃcient, relative change in area under the CDF curve, sample tracking plot, average silhouette width.

97 Appendix B

Chapter 4 Supplementary Figures

98 A Derived HER2 ER PR P53 Lapatinib From Status Status Status Status GI50 Invasive HCC1419 ductal + + + mutation 0.27 µM carcinoma Invasive BT474 ductal + - - mutation 0.40 µM carcinoma B C

HCC1419 BT474

Figure B.1: HER2+ lapatinib sensitive breast cancer cell lines used in this study. Characteristcs (A) and micrographs of HCC1419 (B) and BT474 (C) cell lines.

99 Cell Line Day 1 Day 13 Day 28 Day 40 Day 70 Day 79

HCC1419

BT474

Figure B.2: Drug tolerant expanded persisters begin to grow in 40-70 days. Micrographs of HCC1419 and BT474 cells after treatment with 1 µM lapatinib at various time points.

100 A BT474

C HCC1419 BT474 +L +L24 -L +L +L24 -L H3 H3K4me3 B H3K27me3 HCC1419 H3ac

DTPs (+lapatinib for 9 days) Parental Population (+lapatinib for 24 hours) Parental Population (-lapatinib)

Figure B.3: DTPs have a distinct chromatin state. A,B: Quantiﬁcation of im- munoblots shown in (C) using ImageJ [123] for H3K4 tri-methylation (an open chromatin mark), H3K27 tri-methylation (a closed chromatin mark), and H3 total acetylation (an open chromatin mark), normalized to total H3, for DTPs (green), parental population treated for 24 hours (light purple) and untreated parental population (dark purple) in BT474 cells (A) and HCC1419 cells (B) reveals a slightly more closed chromatin state in resistant cells. C: Imunoblots probed with antibodies against H3, H3K4 tri-methylation, H3K27 tri-methylation or H3 total acetylation in DTPs (+L), parental population treated for 24 hours (+L24) and untreated parental population (-L) in HCC1419 and BT474 cells.

101 FAM21BTNS1 LOC619207AK097527PSTPIP2CCDC92HS6ST1PTPRUMEF2BRGS11ABTB1SIAH2PISD LOC283050AX747261KIAA0195CAMK2BTFAP2AC5orf64AMY2BCNTLNFOXN3MLXIPJUN MCF2LLOC283663PRICKLE1AX747158ZCCHC14ZFP36L2BCORL1SETD1BNT5DC3MUC19KIF26B−AS1 LOC400958BC042385AX748268AK128400AK095700CYP2G1PSLC16A4NRCAMGPERCAP2TXK LOC440905LOC150776HOMER2STXBP1PCBP1ASAP1PHF16KIF5CRCN3GAB2IRS1 ATP6V0A4AX748239C20orf112C21orf15GABBR1TOP1MTPABPC1PABPC4PCBP2LRP12SACS ANKRD20A11PANKRD20A2CYP4F35PAK127460AF339817KIAA1107EXOSC3FAM40BRAB3IPPDE3BAPLF LOC100506385LOC100133161RASGEF1BCYP2D7P1AK055254C1orf220FAM41CMACF1PCGF5LRP5LAR LOC731275DQ573539BC040333KIAA1467DTNBP1MS4A10ZFAND4C18orf1CPNE7ATHL1SIK1 AX748265AL832384C10orf82SH3TC2MERTKDNAH5LRRK2PLAC8TNNI1TAF15TRO LOC149134KCNQ1OT1BC090058BC035181TDRD12LRRC69TAS2R5TRIM66MLANAAGAP5HIPK2 BC047364BX641154BX648961BX647715CCDC120FLJ10661FBXO31NOP14MFAP5USP35KSR2 LOC100131257ANP32ABC040563AK056486SORBS1POU5F1SCN11APARP11ATP1B4HOXC6LRRC2−IT1 LOC100133091RUNDC2CBC019672AX748195TMEM236ZC3H12DSLC14A2WNT9BBMP8APRR19IDUA BC041007AK093203AX748283TMEM136FLJ36777FAM86B1PRR23CNPHS1ZFHX3EP400DMC1 NCRNA00269SND1KIAA0922PODXL2DUSP16LY6G5BGLOD5STAC2TCF25AASSDEF8−IT1 LOC100293516LOC79015BC160930AK094692AX746599ALDH1B1POLR2AASTN2ZMAT3HES2SIAE BC148242AX747521PCDHB6SCARB1ZNF215C9orf72CARD6GFRALTUBB3CD209ITGA1 CACNA2D2AX746635EP400NLUNC13AGPRIN1CIRH1AHOXC5KIF26AEFHD2PVT1MEI1 Color Key BC013821AK126380ZNF618ATP8B2ABCC8PVRL3TACC2CRIM1NOB1SDK1STAR CCDC88CPTGER3ARPC1BCYP2D6ODZ3 LOC100505933LOC283767AK302511SLC12A7SPTBN5FBRSL1SLC9B2FBXL16ACPL2CMIPMLL2 BC045578BC051736CYP2E1ZNF814C9orf89MEGF6KDM4BFOXO1GRB14PDPRMNT LOC728323CCDC165KCNMB3DNAJC6DCAF10PDCD4HEBP1IL6STEAF2LRP4EBF4 CLRN1BC045192AX747080DNAJC22SEC14L5SLC15A2JAKMIP3CORO6PLCD3QDPR−MYCAS1 BC053669C10orf113RNF144APABPC1LSLC23A2C6orf141CCDC78KCNK15PMP22BICD1TRH LOC100507117LOC100294362ASMTLBC026354RAMP1CERS1CBLN2EFR3BCLYBLCOLQPRKX−AS1 LOC100129617BC045779AK125684FAM95B1MYO15APCDH19TMOD2RGAG4PODXLGRIK4IGF1 ADAMTS17DQ592442AX721280TMEM101PITPNM2CBFA2T3C10orf10BZRAP1PREX1KLF13KAL1 ANKRD20A4BC043227BC029255AK094352RNASET2PTPLAD2TNFSF14PTCHD2FADS3EMX1FGD3 LOC285972BC032415FBXW4P1AK056556ZCCHC24GAL3ST4SLC13A5PCYT1BFAM78AAPBA2IDO2 BC024027MAGEA10AK125769AX746575AX747592DNM1P35FDPSL2ASLC36A3X05128BRD2PAPL LOC100507424LOC100287225BC035084CYP46A1MCART6CDH23TEX11ZFP42VSIG1STAT4AQP6 BC034929BC035187FAM123CIGF2BP3KRT42PZNF865WFDC8DUOX1SALL4MILR1SCXB SERPINE1AX748371AX746834ZBTB20CABP4KCNJ5CD3GFCAROTX1CCL5FUT2 LINC00485ADAMTS4AK127190FAM155BRNF207FBLIM1ARGFXDACT3PAPLNMBD6NTN5 LOC100506688C14orf105BTBD16DCDC5BHMT2PTCRADAGLABAG3SV2AZIC2SAG PABPC1P2AK131520AK096230FLJ42102CD300LGRNF125AGBL4DGKBTGM2CD96ME2 DENND2ARASGRP3AK125437BX641143AX746787AX747721B3GALT5FBXL13PDE6APDZD4PARK2 LOC728190LOC440300AK056817FLJ14107EPHA10C5orf45C4orf26GALNSZMIZ2MLL4TLE3 LOC644215HSD17B13DQ595103BX537710PCDHB18CACNG8ENTPD1NYAP2MEFVIAPPRRH Metazoa_SRPLOC285103BC037540BC043213AK126253KIAA2012HAVCR2EIF5A2RTDR1HCN2TSKU LOC100190938LOC284260ADAMTSL2HLABC036850AX746871AK124523BX648496EMX2OSSEC31BLAMA1−DRB5 FBXL19MIR143HGAX747182AX747333AK123617AK055666FLJ42627C17orf99CCDC40AS3MT−AS1AXL LOC100190939SLC22A20BC016143AK022382TMEM212RNASE10KBTBD12FAM157AWDR31ATG9BF8 LOC283089BC043205AX747648AK125858AX747756MGC4836KIAA0182PCDHB9DDX10SPEGGATS LOC100616668LOC100132247GOLGA8BAX747586ZSCAN12DNAAF1ZNF677AGAP3AGAP9CSADRFX3 ASAP1AK096475SDHAP2PARD3BABCC5GSTT2MYRIPCD163SMA4ART1B7H6−IT1 PCDHGC5AL833150KIAA1530KIAA1875FLJ43663L3MBTL1CBFA2T2CNNM3CUX1CCNIIRX5 AX746699AX746755FAM122CFAM193APOLR2J4ATXN7L1CCDC30TBC1D3C3orf62ZBED1ACSF3 FAM160B2SAMD4BMCART1SNAPC4PLAGL2KLHL21WDR81GDPD5ZC3H4HSF1SZT2 BC034268TSNARE1FLJ45513C12orf51R3HDM2ZBTB8AGIGYF1ARID3BBCL9ATF7TAF3 CACNA2D3AK128346PCDHB13AL137655KIAA1549ZFYVE28CHRNB1UNC5BKCNE4ACTG1SCAF4 LOC646762AK123177COLEC12SLC22A4CELSR1LRRC27C5orf42MYD88CHTF8SHC2SF1 ARHGEF3AX748196PCDHB12KIAA2018PCDH17LARP4BHIF1ANPXDC1CPT1AIMPG2AOC2 AK057887METRNLIL17RDPDGFAADCY9PGAP2DUSP8NIPBLSBK1IRS2IRX2 BC047625BC047626BC035179FAM153BFAM215AMICALCLSLC29A3C22orf25POU2F3HIVEP3FOXC1 LOC100652768BC071667AX747507FLJ32756UTP14CZNF208HOXC9DSG3ERN1IFIT3BSN LOC100131551ARHGAP32CATSPER2AK098263AK075019FLJ00310FLJ12825COX6B2GPR83OR2A7LAIR1 ARHGAP28BC035176BC022892AK123543C21orf62SCARF1HOXC10HMGA2VWA2NXNNEB HNF1ALOC283693ARHGEF4BC031979AK027091AX746504AX748181UNQ6975ERVVNMUR1NUAK2−AS1−1 LOC399753AK124578AK056253KIAA1751HP09025DPRXP4MCF2L2CNTN4NDST3PPEF2LYN LOC100134713LOC644961TMEM213POU2AF1SUGT1P3MYO15BNPIPL3SMAD9SNX29ATCAYSV2B LOC100287314AK124439GPR172BSTAG3L1POLR2MB3GNT6ACCN3CHST1PTMSTPP1NFIX LOC100133331LOC643988AK091729FLJ14186TBC1D17U2AF1L4CASP10SLC4A1ZNF598ZFPM1SHOX cytochrome_bSNORA74BKCTD13ZBTB7AMYADMZNF580ZNF296REXO1MLLT1EPN1EMD LOC100128682LOC613037LINC00174AL832007FAM127BNFATC1ZNF703WDR24FOXP4NWD1PSCA DKFZp434J0226SNORA74AARHGAP33BC048427AK296966RTN4RL2SAMD4AINHBBZFP36RBP1SRM MCOLN1CACNG4TPTEP1MED25MYH14RHOGORAI1TRIM8PCIF1BBC3ABL1 ATP6V0A1ZC3H12ASLC25A6SLC2A1GPR56GPC1JUNDKAZNSOX6F8A3CTIF ADAMTS19SLC25A35AK023304PECAM1C20orf54NLRP12TCF7L1CEBPDINPP5JSULF2HIF3A DKFZp434P0216BC045769AK023515GALNTL2ZC3H12BCYP1A2TRIM58CHST6BCL3RHDKLB LOC386597SIGLEC10BC036311KIAA1614ALDH3A1VWA5B1IL12RB1NLRC3SCIMPLUSTINMT LOC100506241LOC100190940LOC731424LOC23117FLJ12334MIR3648TRPM3ASB11EGR4CD52LIPC LOC100130331MTRNR2L3MTRNR2L6MTRNR2L5BC045805AX746877GPR65CCL16ACTA1EXD1NPIP LOC145663PRKAR1BAK125412AK096249MLXIPLPRKCGMYH7BP2RX5CDYL2PRNDCHD5 LOC387723LOC151475BD495725BX161428ZDHHC11FBXO40SH3D21PIK3CDZNF662PRKCASTRC APOBEC3FAX746771AX746617CLEC18BFAM71F2SLC7A14SLC13A3ZNF689C2orf83C7orf46ABCB5 UBE2Q2P1BC038742BC040646AX747766AK095045SLC26A8DFNB31SPON2PDE6BTCL1BXKR5 LOC728558AK123816TNFRSF9FLJ40194TSPAN10TBXA2RTSGA13PDE4CMPP4JAK3SYK CEACAM22PTMPRSS12EGFEM1PBC041025AK128777AK127026NCKAP5LTNFSF15ZNF414ZNF835PAK4 GTF2IRD2P1LOC652276AX746851AK055876SRGAP1FAM83HSGSM1TRPM6VENTXTAPBPSIN3B LOC100131096ADAMTS13SLC25A45AX746839FAM71F1SLC24A4VWA5B2MAP7D3ZNF827GNG4LRIT3 LOC100128531SHANK2AX748420AK097289AK095583CNNM2BPIFB1IFITM3ANO7TPPP−IER2AS3 LOC648987LINC00092AK128559BX537921AK024248AK130294ANKLE1MEGF8GPR82ABCA3PTPRJ LOC100507178LOC100130581LOC145820BC006435CAMK2AANKS4BUBAP2LC1orf86CCBE1UPK1BFHIT LOC100507173LOC100506462OK/SWDQ574721BC045784AK131347AX747140C17orf63C19orf55−CD84cl.16WIZ AX746964KIAA1522KIAA0913FAM102ACASKIN1PIP5K1CUBQLN2C2CD2LPKD1P1MOB3APACS1 LOC100287722LOC221442SIGLEC8MICALL1CSNK1DPNPLA7WDFY2SRCAPNCOA6WASF1SSBP3 LOC646324ARHGAP39AX747444BCL2L15C6orf132PCDHB3DLGAP4ATXN2LWNT9ARNPC3SDC3 AK096219EPS15L1C16orf55SORBS3ZDHHC8ZC3H18GATSL2CDHR4CRTC1RFX1SP2 LOC100289019AX746765AX747238USHBP1SETD1ARAB11BKHSRPMUC20DOT1LSOX12ULK1 SHROOM2BC042590BC062350C17orf70PRRC2BZNF142ASMTLFOXK1SYBUFAT3SPN DENND4BBC045560BC108660FLJ00285C4orf44GFOD1TERF2NCDNPER1KLC2SYT7 ANKRD13BLINC00311HEATR7ACASKIN2CELSR2MAP1SHCFC1RNF44RNF26PALMCIC LINC00202CAMTA2PLXNA1PTPN23NRBP2LENG8AP2A2BCL9LGATA3BRD4SKI C14orf132TMEM201PLXNA3ZNF628ISYNA1CHERPCYTH3PRR12OSMRFN3KATN1 LOC399744SYNGAP1AK098727GLTSCR1C10orf2NUMBLKCNC3NPRL3PANX2SYT15ITIH4 ZDHHC15FAM110BSLC27A4A4GALTZNF773SRRM2NCOR2KDM6BTHSD1SNX22ZMIZ1 LOC649305AX746734FAM176BSLC39A4SPOCK2POM121KCNJ11SSPOESPNCRATCD97 SLC16A1MICAL3DOC2AKIF18BMYLK3COTL1TLCD2SOX13TTC28KAT2APXDN PPANHSD11B1LFLJ33360−CPAMD8GPR135P2RY11FER1L4HSPG2TFAP4ARTNGLI3CFI LOC100129269CLEC11ACLEC18APRRC2ARAVER1PDDC1ZNRF1PLBD2PVRL1RLTPRMSH5 BC045731AX747730FAM123BJAKMIP1R3HDM1PARD6GSTK32CVARSPUS1CADBOK POM121L10PPRICKLE3AX746995RHBDF2FGFR4PRR22RAB42ASB13XYLBTBX1SRL LOC100287792AF090939FAM129CC19orf12DNAH10DCHS1FHAD1FGFR2ITGAXCMBLBRDT LOC148709SLC29A2MIR663AMIR663BSIRPB2FCGBPDUSP2TMTC1IGSF9ABP1ACE APOBEC3DSLC25A34C1orf116C1orf130SHANK2ST8SIA4ADHFE1DAND5CCL22MPLMLL SORCS2POTEMKCNB1POTEFTHSD4PPM1JNR2E3SYTL4GRPRRXRABLNK LOC144481POTEKPPHLDB1ZBTB46SOCS2CNTD2GPR68NXNL2ABTB2IGF1RSGK3 RUNDC3BAK022914BX648270AGXT2L1C2orf27ALRTOMTPRKCEBMP2KRAB30P775PHELB AF079515CDKN1CHMOX1SHOC2ACACBHOXC4RBM20NXPH3PALLDPGM5DLC1 METTL21ASLC25A30BHLHE40TSC22D4D2HGDHSHKBP1PTPN18NKAIN1ABCB8PHRF1GNB2 SMARCA4CCDC124TRMT61ASLC24A6MOSPD3CLASRPPTBP1LSM10MIER2EXD3LDB1 PPP1R13LPCDHB10AKT1S1DAB2IPSAMD1RHOT2MED24NOC4LSPPL3TCF3GRN SLC25A29GHRLOS2TMEM131MARCH8C17orf56TATDN2PITPNAKLHL36CLCN7PI4KAKIFC2 FAM160A2PLEKHA7GATAD2BZNF512BNEURL4FAM53BTOLLIPMED12SF3A1WDR6RFNG TMEM63BAX747325PLEKHN1DENND3CCDC64GEMIN4PIEZO1CARM1TTLL11UBAP2CIZ1 TTC28HMGXB3HECTD3MBOAT7MAPK7ACOX3FNBP4MAP4BRF1−KRI1LSRAS1 SLC26A11SNRNP70KIAA1324MCM3APMBTPS1SUPT5HCLSTN1EGFL7ASXL1CUL9PLD2 CDC42BPBCROCCP2PPP2R2DSPTAN1CTDP1VPS4ALLGL2SIRT7UBR4ANK3DSP SERPINB6TUBGCP6NBEAL2IQSEC1FBXW4ARVCFCYTH1LETM1HSPA5ATP7BIRGQ DQ576756GRAMD4C10orf26PTPN14EIF2C1KLRG2UCKL1INTS3CLIP1BAG6AKT2 AMDHD2SULT2B1ATXN7L2C1orf212ANKFY1WHSC2SMAD7MED26ZZEF1TLN1HGS MARCKSL1LINC00085KIAA1609DROSHASTAG3L2MGAT4BPRR14LTMCC1GNL3LBICD2SSH2 LOC387647HNRNPA0GALNTL1BCL2L11MMP15AGAP1ACAA1EFNB1PRPF8UBE2IPOGZ LOC100093631SERINC5SDCBP2GSTM3PI4K2AOTUD1SNX30RAI14PGS1BAG1UBP1 TXNDC16FAM131ATCP11L2LRSAM1NPDC1SIDT2MSH4SCAPZXDBMZF1MTA1 GALNT10FBXO32MINPP1KCNG1DEGS2SBNO2PDPK1CASZ1ADAT1CERKGLG1 TBC1D22AFAM116BMAN2B2CCDC57HUWE1HEXDCRREB1USP20LTBP3EVPLJPX MAP3K11PLXNB1VPS37BZBTB42ZNF777NISCHPLIN1DVL3E4F1PXNFUK PRKCSHZDHHC3SPPL2BZNF517SGSM2NDST1PDE8BOBSL1H6PDPLECKLF5 LINC00472FAM161BSLC40A1AMIGO1TSGA10FAM43AZNF646STK38FRAT1C9orf3RTN2 PLEKHM1PPLEKHM1SCAMP4MAP7D1ABHD4USP21SH2B1LRP10LMF1TNK1KLF4 TGFBRAP1PRIC285CCDC97TRIM62INPP5KC7orf43OTUD5ATG2AATG9ATMC4NOL6 LOC100129550ARHGAP26CABLES1CYP20A1ZFYVE26ZNF552ARRB1HDAC5ITGA3GRM4NADK CRAMP1LRAPGEF1CLEC16APLA2G10RNF223CEP104WDTC1RAB7AGGA3NXF1ETV3 UCKL1PPP1R15AC17orf103KIAA0247SH3GLB2B4GALT1SCNN1GSEMA4CRPRD2EFNA1IP6K1−AS1 POLR2J3CACNG1NBPF11ARID5ACYFIP2TMPPECAPN1GPR21PEAK1BAZ2AZXDC CALCOCO1GUSBP11KIAA0430SLC35E2C12orf35OTUD7BZNF192BRWD3PILRBMKL2MID2 CTTNBP2NLAL137535THSD1P1RALGDSGPRC5AMAN2A2ENTPD8FAM13BZFYVE1NACC2TAF4 DKFZp667P0924ANKRD36PCDHB16PIEZO2CADM2RPPH1RFTN1RNF43ATP9AUGCGSOX4 MTRNR2L10MTRNR2L4ProSAPiP1HSD17B14TBC1D16SIPA1L3IQSEC2C2orf18C4orf48USP54SP100 FW339974TMEM175FLJ00322CECR2SF3A2IKZF4DTX2TNK2HSF4PIGRBAI2 GTF2IRD1GEMIN8ZNF496PIK3R2MEF2DABHD2MLST8TRPV1FBRSFBP1PML LOC100527964BC021061AX747742STRADANEDD4LMON1BBCAT2ATF6BGNG7ATAT1ETV6 TMEM184BC21orf56C17orf65SLC9A1ZBTB45LMTK3DMWDTESK1INTS1WFS1PKD1 LOC493754PLEKHM2TBC1D9BTACSTD2C11orf68R3HDM4GAS2L1FBXL18ZNF213SPNS1ZER1 DENND1APHACTR4BLOC1S3C17orf28LRRC56TP53I11ADCK1CRTC2HSPB1CSDAPDK2 LOC643837RAB11FIP3KIAA0284ZNF385ADYRK1BTRPM4MED16BCAR1LRCH4TSC2IRF1 LOC100131193LOC283070FAM127AC19orf54MTSS1ZG16BTANC2LASP1YPEL3CBX6PKN1 FLJ23867C16orf13TMEM80B3GNT7LEPRE1MFSD3LPPR2PEX26FBXL8RFX2TAB1 SH3BGRL3GLTPD1C8orf55AHDC1BCAR3KANK1ABCA7AIM1LUROSNCLNTBCD SLC35A4MFSD12PLXNB2NUCB1PPDPFGSK3ATPCN1ITFG3EVI5LSDF4PLLP FLYWCH1PLEKHH3CCDC106CC2D1AFBXO46PPP6R1TNPO2ZFP41ITPKBEML3GBF1 CDC42BPASLC25A23AK130329NUP214RNF167ZNF579ZNF358RPTORSEPN1SPHK2LMO3 L2HGDHTXNRD3MED12LSEPT10C5orf30SMYD3SAMD8EIF4E3REPS2NMT2XBP1 MANEALCDC14ATTC3P1UNC5CCDCP1SDHCPRNPCASKNCS1PKIAEIF1 LOC100506990HECTD2ZNF761RRAGCNFKBIZPRKG1UBE2BZNF91RHOUSPIN3TMC7 C10orf137FASTKD1METTL8MALAT1TIMM23WDR27TPCN2USP14SNX25NKRFURI1 FRA10AC1BC047513PPFIBP1PPP3CBPGM2L1ZNF439PITPNBZNF737DDHD1CDC16SEC63 PPP2R5EFAM134BPOLR1BSEPHS2TGFBR3VPS26AATP1B3ABCC1HSDL1TCF20CCNK TMEM231CCDC47AHCYL1FAM69BZNF749WDR37PLCG1ZNF74TIAL1DFFBSFI1 A1BGRANBP3LFKBP1ATRIM74TRMT1RNPS1AKAP1TSEN2WIPF2RHOBSPG7−AS1 TAS2R30ZNF350MARS2FOXJ2PUM2GNB1TOB2VAV2MXI1IRF6TTL ANKRD20A9PANKRD20A5PGOLGA8AAK128619SERTAD2N4BP2L1ACADSBLRCH1ALMS1BCL2JAK2 −2 −1 0 1 2 UBE2CBPKIAA0754PPP1R21SYNPO2HS6ST3ZNF708STON2CROTAOC3MSL2ELF1 LOC100271836SYT1 LOC100216546LOC641298ANKRD18AAK309255AX748175ZNF740ZNF248ATP1A1C3orf58BACE1TPD52 LOC100335030RAB11FIP2AX747590ZNF37BPFAM175AADAM22MLLT10C7orf41C9orf93CDK19FBXL4 CSGALNACT2PLEKHD1FAM164AADRA1ASH3BP5DNAJB6RNF115ZNF431STAT5BLUC7LUXS1 GTF2IP1ACVR1BPMEPA1CHST15C12orf5RMND1PRKCDBARX2DCAF4ATE1CBL hCG_1997999SLC25A25HSPBAP1PCDHB5PIP4K2ACEP164IGFBP4SCN8ASUSD4CEP68ESR1 LOC646329AL832439PRKAG2DNMBPZNF676DUSP1CTBP2NBPF1GATA2NFAT5SIK3 LOC100505676LOC100506599LOC100190986MAP1LC3BDQ598910PLEKHH1GNPTABZSWIM6PPTC7TULP4MYOF AK023288FAM208BMAN2A1UBE2D1FAM83BZNF726NEAT1CHD7BTG2FUT8TAB2 TMEM229BFAM108C1ARHGEF7AK055017KIAA1737COL14A1SLC2A13ZNF37AERCC6RBM33MDM4 LOC595101FAM117BMED13LTBC1D9ZNF813ZNF326LARGECPEB2APBB2KAT6BZHX2 LOC100506123DQ587763AF289551DEPTORABCA10MAST2GPR52NAT10CDONSOS1TRIO DL492557UNQ2790C14orf43LONRF2TPRG1LSGMS1MAML1SNED1MCF2LFOSL2USP6 AK125301SDR39U1SEMA6CUAP1L1ARID1BTIPARPCLCN6PDZD8TPBGIKZF5FTX SMARCC2CR936688SLC41A2ZNF384ZNF516NDST2SYNE1ADAP1CPSF1TAOK3TLN2 HNRNPUL2SLC35E2BFAM65ACSPP1ERBB3SAFB2AXIN1SPENSAFBSNX1CTC1 LOC100272216LOC645431ZFAND2ASLC35E3CAMKK2PDZRN3RSC1A1NCOA1MTUS1TTC18MLL5 STON1ST6GALNAC4LOC730091−BC018860PDXDC2PBX537481GTF2A1LP2RY6COMTCRYMAHRRTRIB1 SLC25A22BC040358FAM86EPPPFIBP2C1orf115CARD14CYP4V2UBE2L6HMGA1DNHD1NME4 Value CHCHD10BC040577BAIAP2L1AL360260KCNMA1SHISA5PDGFBNFAM1ITM2CRGL1LPP BC041030CCDC104TSPAN33ATP2C2MMP17FCRLBLAGE3A2LD1CDH3FZD4PC LOC100132077LOC730101DQ596646TRAPPC2C1QTNF6C19orf66MAPK12SH3BP4AUTS2RNF24SYT5 LOC338817ARHGEF37RALGAPA2PPP1R13BST3GAL5ECHDC2C14orf25MMP25LRRK1ORAI2IL12A RNF138P1AK000470AK095151PPP3CAADPGKFREM2PPARAACSL6PKP2PTK7MSI2 RHOBTB2AL157440CCDC150GALNT6IMMP2LZNF850PRKD1SCFD2AIFM1CCNFNRK CCDC85BNR2C2APCHTF18CLSPNMYO19TONSLRPL13BNIP3UCK2RPS2NLE1 MFSD10ANP32BRPL27APRKDCACACAQPCTLPOLD1PTPN3HYAL3LMF2RPL8 SLC25A1TOMM40UBE2E2DNMT1MRPL4MZT2BNR2F6MDC1MPSTAARSGPI KIAA0930SNAPC2MAP3K8PTDSS2FBXO44DIAPH1FRMD8QTRT1RAB26RRS1TP53 DNASE1L2SLC25A10ZDHHC14SIGMAR1CCDC86GTPBP3ATAD3ATRNP1LIMK1APRTBYSL CORO1CMSTO2PTIMM8AZNF556SNHG3RSU1KLC3MVDATICCKBKHK SLC25A43JA040723TPD52L1CGREF1MARC1ACCN2ARL4AADAT2NEIL2NPM3ADM LOC100506660SOWAHCST3GAL3HSPA12ASLC27A5PHLDA1FKBP11RUNX1NOP16FEZ1E2F3 CDK5RAP2KREMEN1AK097283IGFBP2PAQR7RNFT2ATP5DBCAMVWA1CTSZF3 EXOSC5SREBF2PLXNB3RNF187ZNF503GABRDWNT7BACTG2PRRT3PRR24BOP1 PCDHGB8PAK090700AK056524SERINC2C19orf31CREBBPAGAP6MAGI1SF3B4FURINWNK1 BRWD1BC038464BC041449AK055918BX648826ST8SIA1IGFBP5DNAH1SGSHCOBLTSIX−IT2 ZDHHC8P1CROCCP3CAMSAP3AK124832ANKRD11FLJ45340CROCCNPIPL1CNOT3ATXN2EP300 DTX2P1−UPK3BP1LOC284865−LOC90834AK097370AK125489PMS2P11FLJ14346C15orf2RARAEME2MROSPIB DKFZp761E198LOC283922LOC440354LOC389791PCDHAC2SYMPKCAPZBMATN2CD276TAF1CCRX PCDHB19PAK130076HMBOX1IRF2BPLFAM21ASPIRE2TRRAPPACS2WWP2POLNUSP4 GRAMD1ATBC1D3CSLC45A4SLC4A3ERICH1GPSM1CERS4FARP1ITPK1SSH1FXN AX748417TBC1D25SLC4A11ADARB1ABCC10RGS12LRFN4NELFJAG2CIITAHTT MYBBP1AAX748308HDAC10KCNH3DHX37LARP1ILDR1RERESUFURCL1IL11 hCG_1980662ARHGEF10CENPBD1BC037357AK094577AX747598ZNF276USP36STK10RPL28PPIEL LOC100506334LOC389634AK127555AK091996SEC14L1PLXND1LAMC1PIWIL2RRP12NPCRFUS AK096982AK311282AX748330AK128153AX746823TRERF1ZNF577YLPM1NHSL2LRFN3DBN1 LOC284454BC038455SLC12A9MAPRE3BMS1P5SEZ6L2DOCK6QSOX1PCBD2PELP1HR SLC27A1C9orf167PRKACAPI4KAP1FAM83FVAMP2DAPK1SCRIBCHID1SNX8F7 LOC285484AX748385FOXRED2KLHDC8BZNF613MAST1LRRN2EPHB3ESYT1HES1VCL LOC400043AX746638KIAA2022RNF144BSLC19A1CAMK1DMAP2K3RASSF4COL7A1FSTL4RAI1 LOC100506730AK304826ARHGDIAFAM100APCNXL3HPCAL1BTBD2SCAF1LPHN1AGRNTAF6 ATP6V0E2HERC2P7ATP6V0CSLC24A3PCGF2AGAP8SYVN1GPR35ZC3H3LTBP4AIPL1 LOC100505576EPB41L1NOTCH3PTPRN2RNF208TSHZ3TTYH3ZFHX2KLF16TTLL4PRR7 LOC643770ANKRD52TMEM180PIK3C2BADAM8CECR6AGAP4ZNRF3LZTS2IRX4INF2 LOC100130899KIAA1671ENTPD2MTSS1LZNF467GNA12ABCA2PLOD3CLDN3TRAF7CD81 BC033260TMEM198TAMM41GALNT2TRABDRBM38PTPRSDHX33NFKB2IGF2RMIDN TBC1D24ATP13A1FAM83GPNPLA6BAIAP3NPTXRPTPRFARSAMKL1CA14BCR KIAA0664GATAD2AC12orf65RPUSD1RNF166MYO9BPRR14SOLHFHL3GLI4ERF ARHGEF1PLSCR3ARID1AMARK4RHPN1DDX51NAA60MINK1VAC14PTGISSRF LOH12CR1CSNK2A2KIAA0889FGFRL1SAP130MGRN1ZNF219DHX34S1PR2DPH1CBX7 ARHGAP17RNPEPL1FAM117AC15orf39MAP2K7CHST12FAM63AKLHL38DNM2ANO8RELA DQ571357AK123891KIAA0226TSPAN14SEMA3FGMEB1SIGIRRWASF2LAMA5STIM1RHBG TMEM120BNOTCH1AKIRIN2ZNF341SOCS3LMX1BKLF10DIP2CUSF2JAG1MVP DKFZp686L08115PRR5Em:AC005003.4−SLC7A5P2ARHGAP8CD3EAPCHST10DECR2WWOXWWC1AFMIDPOLL CYP4F30PKIAA1958C17orf62SLC5A5NPHP4TESK2NCK2FGD1FLNBGIT1CSK AK310228TMEM102FAM213BUBQLN4MDGA1ADCK3DUSP7TCOF1ITGB5MYH9JUNB PITPNC1MAP3K6WDR59CCDC6FHDC1TTC7BTELO2SMG7MTF1RBFAPFAS B4GALNT4LINC00294AX748411AX747628NSUN5P2LPCAT1PIK3CAMEGF9NOLC1PVRL2PREP ZMYND17ZDHHC23SLC20A1FP15737POLR3AKLHDC4PDCD11HERC4MGAT5LFNGNEBL AK097803FAM86DPKBTBD11SLC37A2PCDHB2FAM21CACAD10ZNF490DOCK5BMP1NEO1 DNAJC27PCDHB11ZKSCAN1REXO1L1POLR1AHTR7P1ZNF273ZNF527MTFMTRASA4USP10−AS1 RAPGEFL1AK023178FAM120AATG16L2POLR3HZNF335CPT1CCDK10EIF5BBZW2PCNT LOC113386CDC42EP4IL17RELTBC1D2RABL2BLRRC24MCCC2MECRINTS9GAACFD ARHGAP35SLC38A10DENND5BMAPKBP1ZNF692C9orf85PNPT1PADI2NOL9DLG5IRX3 ANKRD16CHAMP1PI4KAP2MORC2EHMT2UBE2OMAST3MED15VPS53MLLT6MTG1 TMEM120ASLC39A11C22orf29LDOC1LGCN1L1ZNF649TRIM25EIF4G1DCAF7ZBED4CGN GALNTL4CELSR3WDR90VEGFAFBXO2TEAD2SEPT3LLGL1IPO13KIF1CSMA5 SLC29A1SLC6A9TRIM47DRAM1HDAC4MYBL1GPR77ETNK2GALMNLNFBL LOC642846SLC16A13KATNAL1FCHSD2AHNAK2SESTD1C1orf21PANK1CTPSSTX2HIP1 AX748285MCOLN3CALHM2TUBA1ASLC7A2ZNF165KCNK3FGFR1CLCN5CCR6RELB CRYMLINC00482AK097143JA040725TMEM64ELOVL7ABCC4EEPD1ABCA9MRASHMBS−AS1 LOC283335DQ570835BC014180C10orf58STARD9ZNF239ADCY3FSD1LAMD1GNAZMST4 TP73EHBP1L1SLC36A2SLC26A2POLR1EFBXO41SLC6A6GRM6SORDGAS8NAB2−AS1 LOC284379LOC388796SH3PXD2BSLC15A1ADRBK2ANKS1BZNF486ZNF238TRIM36STMN3LPIN3 CTDNEP1AL832797FAM189BSPATA2LFER1L6SMPD4NLGN2RTN4RACTN1VAT1GHR KREMEN2CRYBG3ASPHD1PTCHD1CRYBB3FAM18AAMPD2DOCK3TEX19CHRDTRIB2 TCRBV14S1AX747795KIAA1804ZNF826PSLC35F3PTP4A3BRSK2LMNAMPP2IFT57PIM1 TMEM105SLCO4A1SDR42E1CORO1AMAGEE1GPR153DPYSL5FRMD3HAGHLKIF3COPTN STAMBPL1NSMCE4AAK000798NSUN5P1KIAA1199MICAL2CRIPAKPNMA1TEAD1NAT8LFZD5 PLEKHM3AX746867POM121CFLJ22184FAM179ASOWAHBLY6G6CTICAM1ZNF587MICAL1IL6R PPARGC1BBC035340AK125257C20orf96PDLIM7KCNK5PPRC1EPHA6CTXN1SYN2EDA FW340097BC020935C14orf49SHANK3MAPK15PDLIM1ABLIM2OPRD1HTR1DHHLA1LCN12 AX748125AFG3L1PFAM124ASLC17A7ATP13A2DBNDD1LRRC3ZNF69CD47LRP2HLX STXBP5LPTPRGOLFM1RBM24MYH15ESRGMAFBEML5NAV1ZIC5IL4R LOC728758PPP2R2CRPS6KA2RPS2P32CKMT1ASELPLGSYNPOMTND5STRF6EIF5AMT1E SERPINB5SLC16A5CD163L1CKMT1BGOLM1FKBP5ACCSCTSCSLIT3BCL6PAM TMEM185AMTRNR2L7AX748135RRPL13LC19orf80NKX3PHKA1ACTC1ACTA2ADD3PAG1−1 SNURFHS3ST3B1HM358993CACNA1HC1orf226−NOS1APIGSF9BSNRPNPTGESSSTR2NR2F1PLCL1 TNFAIP8L1SRD5A2CYP8B1ZNF727ABLIM3DHRS2GREB1PRINSDOK7SUN3SOX9 ZC3HAV1LKCNAB2PDLIM3WDR62GP1BAGRB10ENPP4ENPP5LRFN1FLNATP73 dtp_mg05 AK094963AMOT CEACAM20BC070093BC014606BC043223CYP4F22SPOCK1DIRAS2PAMR1KCNF1GLDCLIPA SLC25A48AX747191KIAA1653CNNM1BPIFB2KCNC1ADCY2MGAT3GRIP1RGS3ABAT LOC100268168AK074469C6orf192ATAD3CCYR61SGCDENC1ITIH6ATF3LIPFFAH AX747537RASL10BSLITRK4C3orf67C3orf80KCNG3FANCADPP10EMID2NREPCDT1 PLEKHF1SLC43A1PRSS23ERVVSLC7A1DNALI1PDZD7TUBB6CBLCPFKPPTRF−2 FW339973DQ593686AK127846GUCY1A2PGLYRP2PABPC3CASQ1TGFB1AXIN2OAS1KLF2 CDC42EP2COL13A1FAM196ASEMA6APMFBP1COL4A5SEL1L3OXCT2VASH1RAB32MAOB AX747826C11orf88PCNXL2OBSCNBMP8BSUSD3S1PR3TBX3SIM2PPIFLIFR LOC100128881SH3BGRLCOL27A1TGFBR2PNPLA3NR4A1EGR1EGR3MATKFOSRET SERPINA3AB074160AK127378MTHFD1LTFCP2L1SEC14L2KCNH1RASA3HSPB8SFXN3MEST ANKRD19PFAM189A2AX747913VSTM2ASEPT11OGDHLPGAP1VWDEMYLKGPC6KIF12 BX649128FAM13CSESN3DGAT2SH2B3EGR2ODZ1OAS3CD22RPA4MAG KHDRBS3CNTNAP2RAPGEF4PLEKHA6OSBPL3CYP4F8HMGN5ARNT2EPHA8ETS2PGR PALM2CCDC88ASLCO3A1UGT3A1FAM20CARPP21−C9orf47AKAP2PARP4ACSL4CST4MKX LOC283174BC033373AK125288SLC36A1GLT25D2CXCL14ATP10DLRRIQ1NEURLODZ4KLF6 LOC344595DIAPH2C2orf66LRRN1CLIP4CDH2DTNASLIT1STC2LAG3AFF3 LOC401074TNFAIP8L3STEAP3TSPYL5PDZK1PVALBRIMS4TIMP3GAS6MBPPYY BC011243AK026965LDLRAD3ADRA2CBAHCC1SUPT3HLDOC1TTYH2SATB2FJX1NFIC AX747684FLJ46481ANTXR2PLXNA4FAM84ASPTLC3RAB3BNOD2PITX1FOSBSCIN DNMBPLOC282997LOC286467BC131768CHRNB4ATP1A4OVGP1IL17RBAPOA1CHDHTMC1−AS1 LOC729737DQ583756CEACAM1AX747440KCNMB4TMSB4YCLSTN2GATSL1IFITM1UCP2OAS2 HIST2H2BASCGB1D2CDCA7LSFMBT2MYO1GTUSC3DDX60MUC2DDR2PDK4IRF9 ARHGAP11BDPY19L2P2UBE2Q2P2AK125429IQGAP2ATP1A2SCML1SPRY2SNX7STC1KLF7 LOC286002SLC26A4MAP4K4UGT3A2ELOVL2MTMR2HOXC8GFRA1GRIK3ANKHHCK AK093600FLJ42709BTBD11CSMD3PRR11THBS1SYTL2FAIM3SOX5AATKCBS HIST1H4KTUBB2APRODHHHIPL2PYCR1CPEB1TUBB1ALPK3SNAI1GARSADM2 LOC100129034CREB3L1C10orf81SLC7A5MGST3MYO3BTHRSPICAM1DOK3NEK6TNS4 ABHD12BC12orf24RECQL4NUPR1NR1D1DYRK4HEG1PKM2MYL9FAT1FN1 CR627135DDX12PMMRN2MEX3APAQR4DDX11FADS2PAICSE2F2ESDZP3 C17orf76MTHFD1ST8SIA6RAD54LZNF714ATP11APLOD2SRPK1DDX21SEH1LHELLS−AS1 NEURL1BANP32EDYNLT3ZNF695ZNF730PPFIA4DHCR7FANCBHPRT1TIMP2LDLR AX747630C11orf82IQGAP3AURKBCXCR4INSIG1TACC3ESPL1KIFC1PLK1PIF1 C17orf53FAM72DFAM64ATROAPCDCA3CDC20SPC24CEP55MKI67AMBPCIT PKMYT1DIAPH3FOXM1PARM1MYBL2IL1RNTGM3TYMSGPC3LAD1FHL2 GALNT7PAPSS2CDCA8DUSP6RGS16TGFB3PPEF1PLEK2ASPHTESCGIPR LOC100505989SCGB2A2LDLRAD2HMGCS2STEAP1TM4SF1LGALS1PHGDHCHAC1ITGB2APOD TNFRSF12ABC016972C12orf34CDC25BSLURP1ADCY5KPNA2ITGA6CBR3FUT3PLD1 LOC729970LINC00152BC015784VSIG10LTSPAN4DUSP4MS4A7PHF19OTORBCHESHC3 PCOLCE2RAPGEF5PPP1R1BHSD17B2CYSLTR1MS4A14LIMCH1S100A6ULBP2TCN2MDFI DQ577420GOLGA7BRASGRF1HIST1H3ISLC7A13S100A14ELOVL5ANXA1mirCOCHITGA5−34 HIST1H2BGHIST1H3HHIST1H3EHIST1H4BMAPRE2AURKAGINS3PGK1SHC4RTTNRPA3 LOC100128191TSPAN15CDC25CAKR1C3RPL39LCCNB1PDE5ACKS1BGDF15H2AFZCO9 ADARB2HRASLS2FAM131BGPCPD1TMSB4XKDELR3LRRC31TMBIM1CRISP3S100A1ITGB6−AS1 LOC100049716RHOBTB1MGC2889FUNDC1PLSCR1STK17BFUCA2ACSL3MDKSYPGK HIST1H2AJHIST1H3FNUP62CLSLC6A11SHCBP1KDELC2ERRFI1DEGS1SPAG5ASNSUAP1 EIF4EBP1RNF157MLF1IPUBE2SNEIL3RND3EME1KIF23EXO1PCK2LY6E HIST1H2BNHIST1H2BHHIST1H2AKCENPMGPSM2SHMT2SGOL1TOP2AHSPA2BTG3TPX2 HIST1H3GHIST2H3CHIST1H3BDLGAP5CCNA2KIF20AUBE2CCENPABIRC5KIF2CKIF14 LOC100133669ERCC6LCKAP2LFAM72BNCAPHCENPKCDCA2HMMRASPMCDK1PBK KIAA1524DEPDC1CENPWNCAPGCENPECDKN3CENPFCENPINUF2BUB1SKA3 AKR1B10NUSAP1CCNB2KIF18ANDC80BUB1BRAD51ANLNPRC1NEK2OIP5 ARHGAP11AFRMD4BTMEM40RHEBL1PHLDA2FAM72ASPAG17RIBC2SCDSP8TTK HIST2H2BFDQ786304KIAA0101ESRRGGCNT3ACOX2SPRY4PADI3GPC5ETV4ETV5 APOBEC3BLOC541471HIST1H2ALHIST1H1DHIST2H4AHIST1H1BHMGB2UBE2TDDIT4ECT2E2F7 HIST1H2BMHIST1H2BOHIST1H2ABHIST1H2BBHIST1H3CHIST1H3ACCDC99TUBA1BC5orf34GINS1ACE2 HIST1H2AGHIST1H2AHHIST1H2BEHIST1H2BLHIST1H2BJHIST1H3DPOU4F3AKR1C2CASC5NUDT1CKS2 HIST1H2BIHIST2H3DHIST1H4ASLC7A11GAS2L3KCNJ18BACH2PTTG1FABP6PSAT1KIF4A HIST1H4FHIST3H3DHCR24SPRED2PCDH20RASD1ACER1CAMPSKA1NAV2TLE6 SRGAP2P2C16orf59MID1IP1FAM46CFAM83DCENPNHJURPGTSE1FABP5DPF1LRP8 RNASEH2ACADPS2SIPA1L2ARMC3CENPPXRCC2BRCA1ACOT9TRAIPMCM6ORC6 NME1S100A10C19orf48DHTKD1CAPN13−ACOT7H2AFXACSF2PSIP1MXD3NME2TUBB TNFRSF10AHIST3H2BBSNORA71BRNU4ATACAX747493SNORA21FAM53CDTYMKRHOAPLP2E2F1 DEPDC1BC15orf42C1orf135MCM10CDCA5TRIP13LMNB1RTKN2RPLP2GINS4POLQ LOC100288637SRGAP2CDCA7CSRP1BRCA2CHEK1RRM2MND1E2F8GGHBLM RAD51AP1RACGAP1C18orf54PARPBPMAD2L1ESCO2KIF20BFANCIKIF15PLK4AEN ZNF367CENPHDSCC1STMN1PSRC1ASF1BGINS2GSG2ORC1CDC6TK1 NCAPG2SNRPD1HMGB3UHRF1RBBP8CDC45GHITMSPC25MELKKIF11DTL SNORA84SNORA75TIMM17AITGB3BPNDUFS4RPL35ASNRPEKCNS3RPS18RPL31MTBP SCARNA23RABGGTBSNORA2ASNORD10SNORA34SNORA48SNORA18CTNNAL1SEC61GHSBP1CBX3 LOC100499405SNORA71DTMEM194ATMEM38BSNORA70CEP57L1ZWILCHWDHD1MIS18ARMI2LIN9 SNORA27SNORA22SNORA54SNORA67SNORA52SNORA10SNORA24HSPA14CKAP2VRK1MZT1 LOC283104SNORD15BSCARNA4SNORA40SNORA38SNORA63TRNA_TyrTRNA_IleSNORD8RPL39HEY1 SNORD97C11orf73COX7A2CALM2RNU11CTSL2EIF4ESHQ1XPOTMLH1CD63 SNORA14BSCARNA18SCARNA1SNORD94SNORA79RTN4IP1TMEM60NDUFB1COX6CTPRKBLSM5 SNORA38BSNORA37SNORA68SNORA28SNORA11SNORA3RANBP6SS18L2SCOCIMPA1CNIH SNORA5ASNORA31SNORA49C16orf87AGSK1CLIC4GEN1MLKLTLR1RDX7SK dtp_mg02 BC132896 SNORA11BRNU6ATACC10orf112MGAT4APRKAA2GPR180SNRPGPGAM4P4HA1UGDHGGCT STARD3NLC13orf27EXOSC8EXOSC9RPS15ATM7SF3ERO1LMDH1BORANAB1TXN ZNFX1SNORD89SNORA20EPB41L2CCDC18FLVCR2WDR76ACBD7CETN3DCLK1SLIRP−AS1 MPHOSPH9LOC401397TRNA_LeuSNARGLUD1RPS12RPS24RPLP0U6atacSMC2RBL1−A1 RPS27ANUDT6CENPLRAD18SFXN1CEP78CDC7FDPSLRR1PNPOAT HIST2H2ABC1orf112TMEM48CCDC34MSMO1C4orf21BARD1POLE2FDFT1DCKAK4 HIST1H2AMHIST1H2AEHIST1H4DGRPEL2TRIM59FIGNL1CCNE2RPS3ASMC4PSPHKIF24 HIST1H2ADXRCC6BP1TXNDC17RAD51CCEP152PRPS2RPL36BRIP1RPSAENO1TPI1 TNFRSF10BDARS2KNTC1SASS6ACAT2PRIM1PRIM2MCM8LDHARFC3PDK1 FAM111BCDC25AZNF519ZNF257POC1AMASTLFMNL3ZWINTGMNNATAD2STIL FANCD2DNAJC9CENPQFBXO5TCF19RRM1PCNADHFRCDK2RFC4EZR RANBP1CHAF1ACHAF1BCENPOPGAM1CHEK2LMNB2CYB5APOLA2RPS19LSM4 TMEM191ANCAPD2ZNF732SHMT1MCM5TMPOIMPA2RFC2FASNLIG1EBP SUV39H2TMEM97RPL18AZNF492ATAD5MCM7MCM2MCM4MCM3FEN1DBI LOC100289361NEDD8HIST1H2BKPSMG3FHOD1ALDOA−PRPS1RPS15GM2AMDP1GALEILVBL FAM86HPC9orf140TUBA1CPFKFB4PCP4L1GAPDHPRDX4BLVRBAP2S1ECE2GNE ADORA2BBHLHE41PKHD1L1COMTD1S100A16SLC9A5SLC1A5MARSEREGTRIB3MAFF TMED10P1SLC44A3SLC16A6S100A11SPTSSANT5DC2RPS17LPSMA6TUBG1KLHL2GPT2 LOC100131496SCARNA6C14orf159TMEM139ATP8A2SERHLVLDLRS100PPMELCASTNTN4 LOC100128098TMEM45BNDUFS7SEPW1MARK1PPARGSRPK3BSCL2NPNTPTK6SP5 SNORA59BTNFRSF21ST6GAL1NPLOC4C2orf72ERMP1GSK3BAZGP1RPS29ATP5ESKA2 CDKN2AIPNLMPHOSPH6TMEM107POLR2GUQCRQUQCRHROMO1CMTM1SNRPBDDB2TFF3 HIST1H4HTOMM40LTMEM147CHCHD1CHCHD2FAM86AATP5G3USMG5SNRPFGRHL3ETFB HIST2H2AA3SNORA72CCDC72SLC3A2PSMD5COX7CEPORYARSHAX1IKBIPIARS ANKRD34BSNORA80CCDC167ATP6V0BNDUFA1METTL4INCENPALYREFC1QBPMITFHDX ST6GALNAC5SLC25A20CARHSP1C11orf31ABHD12WDR34RAB20SIVA1A1BGSSR4DTX4 RPS19BP1BC052775TIMM17BC16orf91C19orf10ESRRAPXMP4SEC13TIAF1MTX1AK7 C9orf152C11orf83C9orf142PRSS27AKR1A1HEXIM2ABHD8BMP7B9D1KAT8DDT AK126266GAL3ST2MANSC1CYB5R2SPTBN2SMPD3UNC5ARNF39KRT4LCP1NFIA C21orf119AZGP1P1COL18A1TNFSF10GRAMD2C1orf201PDLIM4MMEL1LOXL1FMO5NDE1 LOC100134229LOC440925LOC401109LGALS3BPST3GAL1CYP1B1C2orf54BAG2KRT7LNP1IDH1 LOC375295LINC00467BC032795C6orf225NCKAP5INPP5DNCALDABCC6BEND6ACOT4FUT9 HIST2H2ACLOC440461C12orf57TSPAN5TMED3DSCR6EPHA1UCHL1TEKT5JDP2PID1 IQCJGALNT14−PARP12SCHIP1ENOX1EGLN3C8orf4MZB1TRILSRCCPEME1 SCARNA20SLC25A12SNORA65FLJ46906C21orf33BMPR1BTMEM9TNNT1SYPL1GPX1AGA DSCAML1ANXA2P1OSBPL10SPTSSBFAM65CHMGCLCNTD1HLASTX19ELL2FYB−C TMEM106ACYP2B7P1DL492607DNAJC17CDKN2AHOXC11OSBPL5GPR110RAB34MUC1CD44 LOC100130855AX747104CCDC74ASLC24A1C9orf103ACTR3COSBPL7ZNF396ZNF655SGCGDOK4 FBXO43TBC1D1ELOVL6GXYLT2C8orf85CSRP2CDH17ARAP1TAF1AH19ID3 SLC16A14SLC16A10ANKRD37FAM198BC12orf32MDGA2GDPD1GSTO2NXPH1EPONES ONECUT2GUCY1A3HSD17B6PYCARDEMILIN1CGNL1DDX25TCEA3TFCP2ACPPDAB2 LINC00256AARHGAP18BC101234AK095117RASL11BFAM105AMLLT11PPCDCRCCD1ARL4DLXN LOC401093KAZALD1ZNF512C1orf56PRDM5ROBO1NSUN7IL27RAPCBD1LRIG3TPK1 EDARADDMETTL7APDE11APARP10MOCS1PDZD2LPHN2PLBD1KRT23BEX4MGP LOC100506810ERVMER34TMEM194BIRAK1BP1AK056897GCHFRMSRB3ALDOCGPR19ELK3SFN−1 LGALS8LOC285419LINC00312HIST1H4LC8orf75LMCD1TRAF1SSPNAPOF−RINLHFEAS1 TRIM6B4GALNT2AK056683SLC39A8−STEAP4ENDOUTRIM34ADPRHBEAN1ENPP3KRT81VCAN AX748345GALNTL6SPATS2LSRD5A3TUBAL3HOXB2RGS10TPM1RFX6KMOPIP BC039389UGT2B10IFITM10PKHD1CASP4LOXL4ECM1TNS3JAM2LHX8EFS HLAAK125397SDR16C5CN5H6.4GPR39TGFBIRHOH−KYNUUTRNCDH5NEU4DMA CDC42EP5SLC16A3CRABP2ABCA12COX8AGULP1PNMTSYT8HGDENGTH PPP1R14BDQ574505AK309286HRASLSMGST2HSH2DFKBP2ECT2LRNLSCD68IFI30 SCARNA12AK094156RUNX1T1C15orf34SERHL2GPR160LGALS8AJUBAIDH2B2MTK2 MTHFD2B3GNT3PTPREHOXB3TPI1P2VAMP5ENPP1LAMB1MDFICCTSHCPM RARRES3HSD17B1SPOCD1C12orf61SEMA4AHECW2COPZ2SYDE1UPK1AOXCT1NPAS2 AK126744HIST1H4IFAM129AFAM100BCACNB4EFEMP1BTN3A1BANK1FSTL1GATMMT2A LOC100506783HIST1H2BFSLC25A33SLC12A8C17orf82RAB27BSQRDLCNTRLMPZL3SDC2PLS1 ATP6V1E2SNRNP27AGPHD1CYBRD1CDC14BBTN2A2IQGAP1GABRPTCTE3RPL2BIER3 PYROXD2PHLDA3RNF181PDLIM5C11orf1GRHL1KRT15FSTL5RHOVRARBPBX3 HIST1H4JFLJ33065GABRG3SCUBE3ATP2B4NEGR1MEIS1RORCSYNCELF5CA8 B3GALNT1RNASELNUDT12MRPL39RNF19ARAB11AC9orf64PRKCHGANCBBIP1U6 SCARNA22SNORA50TMEM182CLDN12MORN2C7orf58FBXW7RFESDSKAP2SAT1PKIB BC034827AL832163SLC4A10ZNF204PPSMB8MAPK6EPHA7GFPT1TPST1AGPSTBX4 LOC100129794DNAJC15MAGEH1SH3TC1C3orf14CABYRPLCH1TEX15ETAA1GYG2LEO1 ARHGEF6GUCY1B3AX747981CCDC19MB21D1TMCO4NOP10ACOT2CLGNTFPI2DLX1 LOC100272217RAD21PPARGC1AANKRD34AHEATR4IL20RAPTH2RHCG4TDO2−CIB3EHFAS1 LOC100506013CR936711TMPRSS4MUM1L1OTOGLHOXA3FAM5CUSP51CRIP1XKRXSLIT2 LOC254559SCARNA14AK056252RAPGEF3SLCO4C1C2CD4ATMSB10C6orf99RORBLIX1LTNIK CCDC109BHSD17B12CHRNA5MFSD2AS100A9WDR67MEF2CS1PR5WSB2CBFBETFA BX538254TMEM154SLC16A9PHLDB2FAM65BSPINK8PDE4DPDE4BPGM1GLRXASS1 ARHGEF25AK056098HOXA10NEDD4SPAG4HYLS1SNX10SNCGPON2CGATIFA TMSB15AATPBD4SRBD1LACC1ASCL4DGKAMYT1PSD3LCA5IQCHNFIB BC039356SCAND3ZNF678PRKD3ACSS3PEG10CDK14HNMTSMN2ARG2ACP1 ANKRD30BAK126307ZCCHC11MAGED1ECHDC1ELAVL2EPHA4LPAR6TIAM1FSIP1ODZ2 C15orf23SEC23ASGOL2DLEU2COQ3HOPXLYZL2QPCTCD36SELLRELT AK127609ARHGDIGCLEC3ATBKBP1MCCD1MUCL1KCND2PALMDGBP2VDRIFI6 par_mg03 SPNS2 AKR7A2P1ADAMTS1AK027179HIST1H3JCCDC160FAM83ATCEAL8SLC1A4AGR2MT1ABST2 ABCC11COL9A3AKR7A3MGST1CAPN8LRTM2MYCNOSR1PGK2ARSJIGF2 BC034795FLJ34503DNAJA4TOB2P1ZNF608GNG12P2RX2MSMBSGCEISG20SGK1 TMPRSS13CEACAM7SLC12A1PDGFDGNA14RAB19ERP27ALPK1SCGNCNN3MT1F HSP90AB4PANKRD30ABC046476ARHGDIBMARCKSMAPK13RASSF6GPR126ABCA4EPHA2RAB31 KRT18P55AX747192ZNF804ASULT1C4MECOMPTPN21ACSM1KRT13VAV3KLF8HDC LOC100289211KIAA0040CYP39A1PITPNM1TMEM18ENTPD3GPR162H2AFY2ZNF750PRRX2BDH2 LOC100289255EXOC3L4FAM167ATHAP10LGALSLTFAP2BKCNU1BFSP1RHOQAGR3DKK1 LOC100130992LOC100506779TNFRSF18SLC22A17PRKAR2BANXA2P3FLJ27352TRIM46SGPP1CCL28TFF1 ADORA2AC15orf48LRRC8CAMIGO2CLDN23CYP4Z1KLHL13HDAC9CDH10KIFC3GLUL TNFRSF10CCYP4Z2PDNAJB4MBNL2VEZF1LPAR3FNIP2FSIP2MGLLSDC1PSG9 LOC100270746C16orf80PLXDC2CYP4X1PSMB7STRN3MAGIXTPM4MT1XDAD1VAPA HIST1H4EABCC6P2CDKN2CHDAC11ACSM3ARL4CBOLA3COX17MAL2G2E3CFL2 TNFSF12−HIST1H4CBC027846TNFSF13GABBR2B3GNT2LRFN2VSIG2SNPHWIPI1GNB4ALPL KRT8P41CUEDC1C4orf38L32131SCN1BREEP1FBLN2KRT80IL1R1SIX2MAF ADAMTS15FAM113BSECTM1PHF21BATRNL1CCL3L1SHISA2PLCE1TACC1P704PFMN1 GOLGA4PCDH10ZNF266ZNF175C1orf63IFNAR1DHX29EPAS1ZFP14NET1AUH TMEM106BFAM179BTOPORSC12orf26RB1CC1TRAFD1UBE2J1ZNF271SMAD4TSTD2TTC14 MRFAP1L1TMEM50AMTERFD3SLC35B3CORO2ATMEM9BSNORA8ZNF624RNFT1PNISROFD1 C14orf101MUDENGFAM135ASACM1LPNPLA8LHFPL2ABCA5NR1D2PSEN1AKAP5SLU7 TMEM150CGOLGA6L9FAM172ASLC38A1PCMTD1PCMTD2CCNG2RSBN1ATG14ORC4SKIL DKFZp451F083LAPTM4ACHMP1BTNFAIP2SCPEP1KLHDC2C6orf47EIF2S3BPNT1LARP6BBS2 SEPSECSKIAA1704RPS6KA5PDCD10GPR98PREPLSAP18ZNF45SPTBASB7BIN1 TRAPPC11PLEKHA8ANKMY2PCYOX1DHFRL1VPS13CSNX13ZNF12PEX1UFL1NBN DYNC2LI1ZNF280DN6AMT1PPP6R3TMBIM4ZNF638ZNF287CASC4LCORLUBE4ASRP14 MPHOSPH10FAM82A2MESDC2SUCLG2NMNAT2STAMBPATF7IP2SUPT7LMTHFSKCNJ3IFT80 ST3GAL4HMG20ADOCK10GIGYF2ZNF187ZNF226C7orf44ZMYM5CASP6PIONNNT ZCWPW1C11orf52ZNF75DMORC3ZNF879ZFP112KHNYNSMAD1CHD9NSFOS9 C20orf177CCDC132ZC3H11AZNF224ZFP106ARFIP1AGGF1NPHP3TSHZ2LRIF1PEX2 BIVMCSRP2BPCCDC126−DYRK1AOSBPL2ZNF181ZNF302ZNF644BRWD1MKRN1ERCC5SETX EPM2AIP1LAMTOR3TSC22D1TSPAN31ZNF518AZNF227ZMYM2TRIP11GTF2BKIF27MIOS TRNAU1APSMARCE1AL355711TSPYL4ZNF304LZTFL1BECN1ASTE1PVRL4CRBNING4 intersectin_1_long_formTMEM184CFAM190ARBM12BRNF13VPS41SSFA2ZNF41PHKBRBL2ITCH ISY1C14orf119RWDD2AGAPVD1GTF3C5−ALS2CLTRIM45PRMT7TTC9CRAB43RYR1IK ARFGAP2TXNDC12CCDC127DDRGK1CTNNA1MPDU1DSCR3DPY30PDCLBRD7USF1 DNAJC25MRPS10RNF141IFNGR1TRIM23NDFIP2CMPK1AMPD3CEP44NFYBPLDN TMEM106CPRKAR1AZBTB41ZNF330PSMG2PNRC2SENP1MIER3NEK7TGFASELT B3GALNT2C6orf130METTL9FTSJD1ZNF322WDFY1GABPAAP1ARVMA21BBS10MIER1 TMEM168LMBRD1EIF2AK3ADAM10VPS33BSLC4A7HCFC2LEMD3SEP15CTBSATG5 TCEAL1INPP4BC3orf38VSIG10NEDD9TMED5NR2F2PAPD5BNIP2XRN2MCU DYNC1LI2PLEKHG3FAM169AMAN1A2ATP11BEIF4G2OSTM1LYRM2PTBP3PGM3LYST TMEM167BAK098012JHDM1DC1orf27RBM18TM2D3FBXO8C1orf9RNF6STK3ST7 LOC100507495LOC338758ZMYM6NBTMEM50BRSPRY1OCIAD1CREG1EXPH5THAP9GUSBITFG1 B4GALT4RANBP2PRMT10CTAGE5ZNF572ZNF277PPWD1COPB2ZBED5SNX14GZF1 TMED7SNORA12MAP3K13COMMD6SLC35A5−SPPL2ATICAM2ZNF134WDR5BMAGT1PAQR8TMF1 THUMPD1AK021570GOLGA5UBE2D3ZNF658ZNF155ZNHIT3UNC50H3F3BSELSTAF2 SLC30A7BCKDHBMRPS14RAB33BZBTB11ZNF133ZNF260ARPC3CAPN7H2AFVZMAT2 C1GALT1C1TMEM30ATMEM39ASUPT4H1TMEM59LYSMD3SPTLC1CMTM6CPNE3COQ7DSTN KIAA0232FAM200BARL6IP6PPP4R1RNF114MYO5AGLRX2VPS4BRNF11VPS45PIN4 FAM104BTMEM66NDUFA8GALNT3PDIK1LHARS2CETN2MFAP3CSTF1COA5ZHX1 BC038570ZCCHC4TXNDC5TM9SF1STYXL1ZNF337HILPDAEBAG9UFSP2HELQGIN1 C22orf28KBTBD4LSM14BSLC9A7LTBP1HEXANBR1HBP1TAP1SIL1VCP ATP6V1E1ZDHHC4ENTPD6PEX11BPET117C2orf47TMCC3PSMD4LEMD2TOR1BMFN2 HSPE1SERPINI1SPTY2D1C14orf80ARL6IP5MAP2K4TM9SF3C8orf76−PCDH9MOB4TPMTCD24 TMEM203KRTCAP2ATG16L1DNAJB1THAP11PRRC1ABCG1SURF4RAB2BMIEN1ATL3 LOC400657COMMD9SLC39A1NDUFB4NDUFS6NDUFA7MRPL52TIMM22WDR5RCN2OAZ1 SAP30BPWDR45LRABEPKPSME1UBE2NPFDN1RNF25CSTF2STAP2ENSAAUP1 BC070371BC073897ZDHHC13FAM102BYEATS4ZNF184ZNF136FOXN2DERANANSEPS8 FLJ23152ARMCX3ZNF844ZNF484FCHO2TRAM1ZCRB1SCYL3RORAFRKGLA CNTNAP3BPPP1R1CSLC30A8UBE2Q2MAP4K3TMTC2NR3C1KCTD6DDX58FRS2CLK1 MPHOSPH8AK130759PTPLAD1N4BP2L2PPM1BHSPH1SREK1SENP7AKAP9FBXL3LNX2 AK311660ZBTB26ZNF197ZNF323RGPD6PDE7ACOG3GCC2DPP8FRYLID2 ANKRD49DYNC2H1FAM214AC15orf29TMEM62SEMA3EGCOM1ZNF770ATG12LTN1TIA1 METTL14SLC11A2C12orf29GABPB1ZNF790DTWD1SMC6EML1IFT88TBK1CAT FAM103A1ATAD2BZNF654ZNF606STK38LZNF449NARG2HSDL2USP12FDX1IGIP PLEKHF2ORMDL1C10orf28BCLAF1ARID4BESCO1CDC5LTOP2BSTAG2CHMABI1 CCDC111RWDD2BHTATIP2ANP32AFAM13AZNF483C7orf23DMXL2SYCP2PKD2C5 GPR75TMEM135C21orf91CCDC14RAD54BFAM69AARL13B−ASB3PIGBINTUEID1AK1 ELMOD2HIST4H4RUVBL1SF3B14TMED2JKAMPSESN1NUP93TAF9BGLO1MMD SNORD15AAK289390RPL36ALSEC11AIER3IP1TTC26POMPSIKE1DCP2OXR1UBA6 OIP5SAMHD1PPP6CPRPF4SPAG1ZBTB6YIPF4RRN3GPD2DLG3TANK−AS1 LOC729852LEPROTL1DQ593252TBC1D23ANKRA2PPP1R2MYL12BFBXO28GSTCDKPNA5SNF8 SLC20A2SEC11CMRPL19POLR3FC5orf44CDC26OSTF1ASF1AHBXIPNGDNYIPF5 BC067244KIAA0895GRAMD3SEC23BTMCO1PSMA2COPB1PLRG1RAD21TTLL7YES1 MRPS18CFAM111APPIAL4FMRPS22FERMT2NUDT21CNOT8FBXO4GALK1RPS11ATP5L SH3BGRL2TMEM123MRPL35C4orf33WDR54POLD3RPL21RPL41PYGLNXT2IMMT par_mg01 GPN3 SCAMP1KBTBD2ZNF410RBMX2CCNT1CEP57MIS12M6PRE2F6PIGASP4 TRAPPC6BDNAJB14C16orf52TROVE2ZNF404ZNF627ZNF189FRMD6CPSF2NAA38ZNF24 SLC25A32DCUN1D1FAM91A1OSGIN2HS2ST1C5orf43UHMK1EXOC5EIF4A2FBXO3ADSS TMEM161BHNRNPH2TWISTNBSLC30A9TXNDC9CAPZA1TATDN1TSNAXUTP23TIPRLNAPG HNRNPH1SLC35A3CCDC43UBE2V2PSMC6PSMA1CEPT1DERL2TTC35HIAT1CCZ1 AF283776AL833455ATP6V1DSERINC1SLC38A2COQ10BRAB22AGTF2F2UBE2WZUFSPMAT2B FBXO22CCP110BNIP3LTHOC7DHRS7SPOPLRDH11PFDN4ACAP2VPS35GCH1 C14orf129AL110181METTL15TRMT11CAPZA2ZFAND1ZNF180ZNF267SRSF7FANCLSEPT7 MMADHCMEIS3P1PTDSS1HRSP12GEMIN2RNF138ASCC3CCT6ACCNCBZW1ALG6 PPP1CCRECQLPRDX3CYB5BHAUS6TRUB1SRSF1CCBL2HPS3VBP1RPE PPPDE1MRPL13C1orf96ARMC1SLMO2EIF2S1SRSF3CYCSSNX5NAE1MLF1 DNAJC19C20orf20PTPN11CNOT7USP37HBS1LUTP15TAF11CBX1ETF1TSN PPP2R5CACTL6ARNF149ATP5F1EIF1AXAPH1BCDC27VDAC3GNAI3TSFMSET SNORA30C12orf75NUP155NUP205SSX2IPNEDD1CSE1LMSH6DNA2DSN1HLTF ANKRD32PLEKHA2LRRCC1SNRPA1XRCC5HAUS2DUS4LTFRCCCT2SKP2RPL9 SNORA14ANDUFA9TOPBP1LRRC40TWSG1NUP37CISD2MDM1PFKMXPO1CCT5 MIS18BP1SNORD9C19orf63FAM60AWDR73ACTR6RPL26ARSDUGP2DEKDUT LOC100506207LOC145837PDE4DIPMAPK10COL9A2ACOT13EEF1A2SCMH1PTPRHHOXA5VIPR1 LOC100288181FAM114A1CCDC134BLOC1S1SLC39A7GOLGA2C6orf165C4orf19ACSS1COPGITGA9 SELENBP1CCDC121ZFP36L1RABEP2PHLDB3TCEB3IGSF3WBP1MAP7PPLID4 BC095475BC150535AX747544SERTAD1ALDH5A1SH3RF1RDH10TCRAVASNLEF1ID1 ARHGAP10DYNLRB2HSD11B2SERINC3C15orf44TTC39ARHPN2PQLC3MCEETHRBNRP2 SLC25A38PTTG1IPPMS2CLDCAF11PEX11ASFT2D3TTC30BMIR23BOTUB1CFLARCRIP2 SLC37A1GOLGA1MBOAT2PSMB10COL5A2C8orf77CAPN9PGAP3RBM23BCAS1C2CD2 C1orf106ZBTB7CC2orf68NIPAL3RRBP1LAMA3IGSF8TYW1GFERNOD1EDN1 TPD52L2GPRC5CAMOTL2EMILIN2ZNF747ABLIM1ABCA1DCXRCLIC3COQ9RIN2 TMPRSS2RAP1GAPC21orf63ACTL10DHRS3CPNE2SPEF1FGF13HINT2GGT1ACP2 PLA2G15C16orf58DNAJC4B3GNT9GPR137CAPN2BCL7CSERF2NAA40GATA6GSN HIST1H1CAF086132FAM162AMYEOV2MB21D2MRPL47TNRC18CEBPBBSPRYMRPL1CHML TMEM242TAX1BP3SLC35F5C18orf32TTC39BTWIST1TMCO3CNTN1RAB15DSC2TTC9 SLC31A2C19orf79SAMD13HECTD1MYL12ADOCK4PLCB3ANO6GRB7GAB1RTN3 CCDC28AOSBPL1AFAM192ACHMP2BSCARB2TXNRD1ZNF830HOMEZPCGF1NPTNISCU SLC7A6OSADIPOR1PRKACBC6orf72PRRG1RAB9ABLVRAUSP25AKTIPGDE1IMP3 ST6GALNAC6MACROD1ARHGAP5KIAA1257HEATR5AC4orf32BAZ1ALYPD3AMFRLPIN2SOS2 LOC100270804UGT2B15FAM211BPFKFB3ZNF532CMAHPSGMS2GSTK1FGFR3TTBK2STOM LOC440335ARHGEF2ARL2BPFAM46BNFKBIAVEGFBCXCR7TM2D1NTAN1SYT12NTN1 SETMARRNF111CCPG1PPM1KHERC6BACH1AKAP6CRYL1DIRC3USP8RTF1 LOC100505648NAALADL2LRRC37A4EFCAB4BCCDC149UGT2B17SRGAP3DOCK11LRRC48ATP8A1NEK10 AK027541POLR3GLNANOS1C12orf23C15orf24GTF2A2DIXDC1NCAM2MUC16DLX4CTH LOC550112PPAPDC1BPPP1R3DSLC10A7ATP13A4C17orf58SLITRK6MAP3K5SNAP23RALBKL SMPDL3ATSPAN9EPCAMPIH1D2AGAP2CES4ARIMS1MLLT3TMC5GBP3SYT9 TUBGCP5ARMCX5EIF2AK4CHST14STARD5ZNF223C3orf52CADM1TIGD4MNX1AQR GTF2IRD2BC047609AK092541KIAA1456SNAPC5C2orf43ARCN1LRP1BUBA5DTX3FAN1 ARMCX5−DENND1BSCARNA9GPRASP2ZFYVE19ELMOD3C11orf71MFAP1N4BP2PDIA4CA2TEF CERCAMCCHCR1C6orf170RASSF7GPKOWSLC6A4PDGFCRNF40FKBP8GAMTGLRB SMARCAD1FAM82BZNF143MYCBPMBNL1PLOD1RHEBADH5NSL1TFPTILF2 ANKRD46HSD17B4UQCRC2NSMCE2NDUFA4KBTBD6FAM96ASRSF10TMTC3TGDSTMX1 FAM49BHMGN1NIF3L1DHX40NUP43NOL11INTS2RARSMTF2SNX4TAF9 SUPT16HTBL1XR1ARPP19ZNF200ZNF510C3orf23BTBD6ZBED6CRLS1APIPTES BC010186CSNK2A1MAP2K1GRPEL1C8orf40NUP54BPGMPGM2ORC3CRYZUBR7 C15orf38RNASEH2CMAGEF1MRPL32MRPL21−SUMO3EEF1DGSTZ1SNX21AP3S2CALRNOL3 RG9MTD2AK093871TMEM223GALNT1PSMC2SHFM1NUDT9CNPY2G3BP2BUD31CNIH3 CCDC71LEHHADHHENMT1SUCLG1CCT6BRAB23IDH3BSOD1IFT43NXT1NVL COMMD4ENOPH1PRRG4PFDN5LIN54CCT7SRP9CCT8RFC1SLBPCINP DQ586540METTL17ADRBK1PPP1CATBC1D8TTC30AOSCP1GPN1FRG1FAF1ADI1 CCDC102ASLC4A1APANAPC10FAM92A1RPUSD2RPS27LZNF230C2orf44MOB1AATP5JASL CYP2U1KANSL2ZNF232C9orf82PGBD2PRPF3DDX23CTPS2THAP6PEX13TTC3 TMPRSS3SCARNA7AB074166RASL11ASLC39A6CNOT1AKR7LSYAP1RBMXLBX2COIL ARHGEF38SCARNA10SCARNA5OSTalphaFAM154BVTCN1CTSDEAPPPSG4CD55RP1 B4GALNT3GRAMD1CCEACAM5C11orf92RHBDL3FILIP1LGLUD2LCMT2ANO1CCR8NAV3 TMEM87ASFTPA2HCAR1EFNB2CLCF1SULF1CRIPTNFIL3UCA1TLE1MYB BC015429ALDH1A1NHLRC3unknownNIPAL2VAMP8ETNK1SNX16FLRT3TC2NPIGF MORF4L1SBDSP1DHRS13LRRC8BCEP120HIGD1AZNF22RHCELHX1TLL1SI TSPAN12RSL24D1TSPAN13ZNF780AB3GALTLDEPDC4ERGIC2TMOD3PSMA4FEM1BDHX15 ALDH7A1COMMD1ABHD10MRPL48C4orf39PAIP2BCASD1LETM2POLE4NIPA1C1D LOC374443LOC645212CTDSPL2SCAPERC6orf168ZNF630MCFD2AAGABUFM1UACAING3 TMEM45AZDHHC20OBFC2AHMGB1MAP1BARAP2RAB8BRUFY3NUPL1USO1BET1 SLC37A3C15orf17CRNKL1KBTBD7DNAJA2YAE1D1USP16SEC62IDH3ACRCPPIGP ZNF518BMRPS31ZFAND6GORABKRCC1PHTF1COG6NIPA2UBR1CLN5LINS ATP6V1G1PRPSAP2NDUFAF1SLC35D1ANKRD5TMEM85N6AMT2NDUFA5IMPACTGCFC2BLZF1 TMEM14BU2SURPALG10BC12orf4MNAT1THAP5AIMP1ALG5VEZTIBTKRIT1 TMEM128EFCAB11PSMD10NECAP1TM9SF2HIBADHZNF268CHMP5EDEM3RAB1ATRAF5 C11orf67C20orf43ZCCHC9TRMT12GTF3C3SUMO2C5orf15WDR61SAR1BSDHDINTS8 DPY19L4MRPS11DGUOKNUBPLHEBP2CASP3THAP3FLOT1FBXL5SBDSSCP2 MRPS18ATRAF3IP2RPS6KB2SULT1A1C19orf53ANKRD9MRPS15PSMC3KRT18UPB1AIP NDUFAF3SOWAHAC9orf163NDUFS8NUDT22MYL6BSRP54ALAS1CHN2CAPSKIF22 dtep_mg07 LOC100507557KIAA0391UQCR10POLR2LPDZD11WARSMED9PPIL3EDF1BAD PACSIN3NDUFS3YJEFN3PSMD8EIF2S2FRRS1PHF17VPS25NENFPLK2PIR LOC81691TMEM205SLC31A1PIK3AP1MRPL40PEX11GKDELR2TOMM7NCEH1MED31TLE4 DCUN1D3C14orf142TIMMDC1FAM173BMETTL6ZNF222C20orf3PSMC1PI4K2BCPEB4HLA−A LOC100630918TMEM99ZSWIM3BRP44LMORC4C5orf54PSMD6P4HA2RPP14TAF13ARF4 MAD2L1BPEXOSC1DHRS7BLMAN2LGTF2H5SLC9A2ZNF295BCAS2SMG8SDF2F11R NDUFAB1RPL26L1AGTRAPSNRPD2NDUFV2NDUFA2MRPL51ALG14MED8SARSATP5I FAM177A1ZCCHC10SDHAF1TOMM6PXMP2ZNRD1PRDX5SPHK1ANXA2PIGUUXT ANAPC11MRPS24DOLPP1MRPL54SNRPCPSMB5COX5BBCAS4RRP36CBX2PHB UBAC2WDR83OSTMEM111C11orf10PSENENATPIF1ETHE1KRT10URM1−H1F0AS1JTB SNORA42FAM110CC1orf123DUSP19SEC61BTRIM69CENPVNUBP1SIRT4LNX1PPIB AK307134DNAJC28MRPL46NT5C3LBAHD1SENP8TCTN2GGCXLMO7PELOCIB1 ANAPC13POU5F1BKIAA1024NFATC4C8orf45C9orf91FUCA1RIMS3NGRNADALDEF6 LOC729013LOC645249PLEKHH2ORMDL2ATP13A5PRKAA1ZC3HC1ABCC3CDK6SNX6LIPH C11orf84CALML5SEC24ASUSD1SCFD1ACYP2PARD3TRPT1LIN7BSRA1BRF2 BC042448FAM176AC21orf59POLR2CSEC22BMPV17LGMPR2C2orf28SCRN1VOPP1PEX12 TMEM208TMEM165ZKSCAN4TRAM1L1C11orf95TMEM56C8orf83DRAM2MED23TUBE1ARL1 PPP2R3CPDXDC1MRPL53ZNF221ROBO2MACC1COPZ1RIMS2STIP1OST4NFU1 NT5C1BUQCR11ARPC5LMRPL20ERLEC1−RBM8ARDH14QPRTIKZF2RPA2PIGVDST CTNNBIP1BCAP31PRR15LGMPPBC5orf35PSMA5OVCA2PHPT1JAGN1PDHXBRK1 SNRNP25UQCRC1C17orf61B3GNT1BOLA2BXRCC6PSMB3POLE3NTHL1GMPSCLIC1 TCTEX1D2UQCRFS1TMEM141C17orf81GTPBP2LYSMD1DNAJA3AHSA1DUS2LTHAP2DRG1 HSP90AB1NUDT16L1THUMPD3HMGN3C3orf75PSMB4SRSF9HSPA9TAF12NMD3CALU CDC42SE1NDUFS1CIAPIN1XRCC4PSMB1UEVLDDCTN6RBBP9STT3ACBR4SELK LOC100294145AK096279SLC35B1ORMDL3MRPL18SUMO1CDC42VDAC1PDHBRPN1LSM6 SLC30A5CD2BP2TRMT1LRNF139DIABLOATPAF1DCTN5SNX11TTC33OXSMTMX2 CACYBPPSMD1MRPL3ACBD3ABCE1PHTF2CDS1ENY2RAE1PPA2PIGX NUCKS1PTPDC1ZNF684ATP5A1NTPCRPCGF6PARP2ALG10ITGAETCP1ICMT ATP13A3MRPS17NDUFB2SFT2D1C14orf2TRIAP1MFSD1SAR1ANAA20PARK7IARS2 COMMD10SCARNA3BLOC1S2COMMD2TMEM65RAP2BLYRM1HIBCHCNIH4NARSPIGK LINC00493TMEM14CZBTB8OSMRPL50COPS8ACTR3YIPF6PPIL1LIN52MYL6PPIC LOC644656MOSPD1NUDCD2MRPS36MRPL22TSEN15C8orf59SRPRBISCA2DPM1LSM3 TMEM167ASLC35A1HOMER1C12orf62TMED10MRPS7TMED1HINT1CAP1TARSUBL5 PSMA7COPS4COX7BFOPNLMGAT2SURF1PHF5AHEXBOPN3LSM1PEX7 DNAJB11PSMD14MRPL27C5orf32C4orf43RNGTTCOPS3MTFR1ATP5HSTK39RPL34 NDUFA12DONSONOGFOD1NDUFS5GEMIN6RPL23AHSPA4LPSMB6RPS20COQ2OSTC RPS6KA3S100A13APOLD1HSPA13UQCRBRPS25BNIP1COQ5EMG1GNB5GBE1 LOC100288748SNORA46AL832909C1GALT1COX6B1IFNGR2HAUS4ACSL1MPP6GPX3POP1 LOC100288428RPL23AP32ALDH1A3METTL18GNPDA2NDUFB8FBXO16TIMM8BLMAN1CASP7RPS26 DNAJC10PGRMC1NDUFB5SNRPB2MRPL33GOLT1BHMGN2RPL27RPL35ERI2PTS TMEM126AMAGOHBFAM18B1ARL6IP1POLR2KTGFBR1SOCS6CEP97C4orf3AZIN1ZAK LOC494150TRMT112NDUFC1NDUFB3TOMM5C6orf57EIF2B3ESRP1MPZL2PLS3NPL HIST1H2BCHIST1H2BDNDUFB6EEF1B2RPL37AC4orf27RPS14APOOCCT4RBX1SAE1 C6orf125NDUFA6MRPL11FAM84BTIMM10PSME2ANXA5RPL37RPL32PECRGSS MTRNR2L8ARHGEF40WASH7PCDKN2BBDKRB2RUSC2OVOL1SEPP1H2AFJFA2HCEL TMEM86AC1QTNF1SLC12A4CDKN1ASLC1A2GDPD3CAB39ITPR1PERPGPX2CLU LRRC37AKIAA0355STK17AATP2A3DEDD2EFNA3ZNFX1ZNF79DIRC2BNIPLUCK1 MTERFD2TMEM178KIAA1147SLC30A4C5orf41PBXIP1ASAH1RC3H1SPSB1SPOPAKD1 LOC400027ANKRD13ATNFRSF19ANKRD12NPEPPSKLHL24PIK3R3DCDC2EIF2C4ERAP1SWT1 LOC100505783COL28A1FAM19A5SLC44A1PRUNE2ZNF596CMTM4HERC3RIPK4DLL1MB KIDINS220CCDC48ACVR2ATSPAN1RIMKLBCITED4MARK2SSBP2PUS10CIRBPOSR2 CACNA1DVSTM2LLGALS3ERBB4PRSS8RMRPTBX2CA11WTIPELF3TJP1 LRRC37A2KIAA1377FLJ10038LRRFIP2C11orf63SMAD6NRIP3KRT9CLK4DLX3BMF ARHGAP24LRRC37A3MIR600HGLOC90246HERC2P9SLC35C1DNMT3ASCUBE2SEC24DSPINT1SPR FAM13ACOL17A1ZNF853C3orf70CMYA5TRAK1SPEF2SNX33BRD8CRY2−AVILAS1 MTRNR2L1HLALINC00284SLC9A3R2GADD45APCDHB4SCNN1BGPR146−FOXD2EFNA5ATOH8DQB1 HERC2P3DOPEY1SAMD12C6orf97SMAD3KIF13APCSK6RAB6AN4BP3RERGNAT1 LOC645513C16orf45OGFRL1SAMD11MAN1A1CCDC93AMOTL1OSBPL9PIK3IP1EFNB3PLAC2 ATP6V1C2HERC2P4HERC2P2AGPAT2ABCD4FOXO4RGS22P2RX4GRIK2ITGB8HHAT ANKRD50CYP1A1ARNTL2SHISA9ABHD3INSIG2ATXN3TVAS5NRP1INSRSIX4 dtep_mg08 dtep_mgA6 dtep_mg07 par_mg01 par_mg03 dtp_mg02 dtep_mgA6dtp_mg05

dtep_mg08

DTP 1 DTP 2 DTP IK SI F8 F3 F7 KL U6 C5 TH AR PC HR GK MB ID3 ID2 ID4 ID1 SKI CFI CIT PIP AIP DBI PIR CIC GPI IFI6 TTL JTB LYN WIZ SF1 JPX FBL ZP3 TK1 H19 TK2 ST7 CAT AXL KLB SP2 FTX LPP FN1 TTK SP8 DTL 7SK AK4 SP5 AK7 AK1 SP4 PPL TEF ASL FAH OAT JUN TXK SRL MLL LSR CBL HTT IL11 VCL HLX FYB LXN EFS CA8 SET CA2 TES NVL RP1 PTS ZAK NPL CEL SYK FUK FXN FUS ERF SRF NLN RET PYY SYP PBK TXN EZR EBP SFN HFE EHF OS9 NSF GLA FRK IGIP TSN ILF2 C1D UXT DST CLU EDA ACE PAM LIPA TRH ME2 BSN NEB SPN MPL DSP PXN PML KHK CKB CSK CFD FOS CBS ESD CO9 BLM AEN PNP DDT ME1 CPE NES B2M AUH NNT VCP SIL1 RPE DEK DUT CTH BAD PHB SPR SAG AGA TRO TIFA AVIL NXN BOK CAD EIF1 SFI1 NRK CRX BCR GAA IL6R IL4R LIPF HCK PIF1 SCD LIN9 RDX DCK TPI1 STIL HDX SRC EPO VDR MAF NBN TIA1 GSS BMF AQR ATIC SIK1 MNT ZIC2 RRH RHD GLI3 CIZ1 HGS BAI2 IRF1 IRF6 SIK3 INF2 GLI4 MVP ZIC5 LIFR PGR MKX MBP IRF9 LIG1 GNE TRIL ENG CGA HDC SIX2 SKIL ABI1 GSN MYB RIT1 SIX4 IRS1 MYC IGF1 IRX5 IRX2 IRS2 EMD SRM CTIF LIPC IER2 FHIT GRN KRI1 MVD ADM TSIX RAI1 IRX4 GIT1 IRX3 CGN HIP1 GHR PPIF MDK PID1 HGD CPM IER3 CIB3 IGF2 BIN1 PIN4 EID1 ADI1 LINS IBTK CIB1 LIPH ERI2 FAT3 MAG FAT1 ATL3 SIAE IAPP NFIX PKIA URI1 SPIB OIP5 GGH NFIA IDH1 MGP RINL KMO IFI30 IDH2 PKIB TNIK NFIB MCU CHM RIN2 APIP TLL1 PPIB VAT1 MRO LY6E TAF3 ATF7 TAF4 TAF6 ATF3 NFIC MITF ITCH TAF2 INTU TAF9 PISD MEI1 IDO2 IFIT3 NPIP MXI1 MSI2 PIM1 SIM2 FJX1 IARS ELL2 ING4 GIN1 LTN1 IMP3 COIL PIGF ING3 PPIC TJP1 IDUA LYST JAK3 JAK2 FAF1 TAB1 ATE1 TAB2 KAT8 SAT1 TAP1 TLE3 F8A3 E4F1 KLF5 KLF4 MID2 ELF1 TRIO TP53 E2F3 TP73 KLF2 KLF6 SCIN KLF7 E2F2 E2F7 TLE6 E2F1 E2F8 RMI2 TFF3 ELF5 KLF8 TFF1 SELT PIGB MMD PIGA E2F6 HLTF LEF1 ISCU CINP TLE1 PIGP TLE4 PIGV PIGX PIGK DLL1 ELF3 INSR VAV2 VAV3 DVL3 PAPL PAK4 ATN1 CIITA PPA2 NAT1 KAL1 CCNI SZT2 ABL1 INMT ITIH4 TLN1 PIGR PLLP TLN2 FEZ1 JAG2 FHL3 JAG1 ITIH6 AFF3 PLK1 FHL2 GIPR MDFI PLP2 PLK4 CNIH TLR1 JDP2 ELK3 PLS1 RELT SELL CFL2 ALPL UFL1 PION ETF1 LBX2 PLK2 F11R PIGU H1F0 ICMT PLS3 VAPA RTF1 ETFA FAN1 FA2H ABAT RPA4 AATK RPA3 ATG5 RPA2 APLF PVT1 MLL2 CMIP EBF4 LRP4 EAF2 FUT2 MLL4 B7H6 BCL9 TPP1 BCL3 ULK1 SYT7 KLC2 TBX1 LDB1 TCF3 PLD2 AKT2 ETV3 FBP1 ETV6 BCL2 SYT1 FUT8 MLL5 FZD4 SYT5 PTK7 NLE1 RPL8 KLC3 CA14 STX2 FZD5 LRP2 BCL6 TBX3 ETS2 LAD1 PLD1 FUT3 TPX2 ETV5 ETV4 LRP8 RBL1 PTK6 B9D1 LCP1 LNP1 FUT9 TPK1 LHX8 SYT8 TBX4 DLX1 IQCH LCA5 BST2 SLU7 RBL2 STK3 CLK1 LNX2 TBK1 RPL9 TTC9 DLX4 SYT9 TFPT NSL1 TTC3 LHX1 BET1 LNX1 ARL1 UBL5 WTIP CA11 TBX2 DLX3 CLK4 FRYL PAG1 ART1 PFAS LAG3 KRT4 KRT7 KRT9 NAV1 NAV2 JAM2 NAV3 STAR MTA1 TANK TARS TNS1 DEF8 CCL5 SV2A CD96 RFX3 HSF1 SBK1 SV2B LUST CD52 CD84 RFX1 FN3K CD97 XYLB ABP1 DLC1 BRF1 CUL9 IRGQ EVPL TNK1 LMF1 NXF1 HSF4 TNK2 DTX2 ZER1 TSC2 RFX2 SDF4 XBP1 ZHX2 PKP2 LMF2 BYSL NELF RCL1 CD81 MIDN USF2 POLL FLNB CD47 FLNA CD22 CST4 STC2 STC1 TNS4 LDLR SKA3 ECT2 SKA1 DPF1 CD63 LRR1 FEN1 SKA2 ETFB IKBIP DTX4 CD44 BEX4 RFX6 TNS3 CD68 PBX3 LEO1 ASS1 CD36 ARSJ G2E3 NET1 BBS2 ASB7 PEX1 PEX2 MIOS USF1 GZF1 SELS ZHX1 CD24 AUP1 EPS8 FRS2 FDX1 SNF8 YES1 NXT2 IMMT ALG6 VBP1 MLF1 SKP2 DTX3 NXT1 SLBP CD55 CLN5 VEZT ALG5 EDF1 ARF4 SDF2 DEF6 BRF2 SELK TCP1 PEX7 SAE1 OTX1 TOB2 UAP1 DAB2 ADAL RTN2 ACE2 ACP1 ACP2 RTN3 RBFA FASN TGFA OAZ1 GATS PALM CRAT BUB1 HHAT CAP2 KSR2 HES2 SDK1 NTN5 EPN1 RBP1 BBC3 SOX6 EXD1 XKR5 TPPP PER1 PUS1 BLNK HELB EXD3 ANK3 SSH2 UBP1 MZF1 PLEC NOL6 MKL2 SOX4 PKD1 PDK2 PKN1 CBX6 GBF1 EML3 DFFB MSL2 UXS1 ESR1 BTG2 USP6 CTC1 SNX1 SAFB RGL1 IL12A RPS2 CTSZ USP4 SSH1 SNX8 HES1 MKL1 CBX7 RELA JUNB MTF1 NEBL DLG5 NOL9 RELB NAB2 IFT57 SYN2 EML5 SOX9 CDT1 PTRF PFKP PDK4 SNX7 SOX5 NEK6 MYL9 TCN2 PCK2 BTG3 NUF2 NEK2 CKS2 CBX3 MZT1 VRK1 HEY1 LSM5 MLKL NAB1 PDK1 RFC3 RFC4 LSM4 RFC2 ECE2 GPT2 NTN4 HAX1 YARS SSR4 LIX1L PSD3 MT1F DKK1 GLUL MAL2 SPTB IFT80 SETX NEK7 RNF6 TMF1 HBP1 DPP8 IFT88 EML1 PKD2 UBA6 DLG3 CCZ1 HPS3 NAE1 SNX5 CBX1 CCT2 CCT5 NTN1 USP8 RALB UBA5 SNX4 MTF2 NOL3 IFT43 RFC1 CCT8 SRP9 CCT7 TC2N SCP2 UPB1 CBX2 SNX6 SRA1 NFU1 OST4 BRK1 LSM6 RAE1 ENY2 MYL6 LSM3 CAP1 LSM1 RBX1 CCT4 AKD1 ILVBL RYR1 CRY2 DAD1 AQP6 BAG3 VARS BAG6 BAG1 APRT RTTN BAG2 VASN OAS1 OAS3 OAS2 MATK OXR1 ATP5I GAB2 AASS ODZ3 IL6ST FGD3 BRD2 TSKU CUX1 SHC2 ERN1 KAZN JUND SDC3 BRD4 SGK3 UBR4 PGS1 ZXDB GLG1 H6PD FBRS NCS1 TIAL1 SPG7 SOS1 UCK2 RRS1 RSU1 BOP1 DBN1 PRR7 DPH1 FGD1 NCK2 LFNG CTPS MST4 GAS8 MT1E SLIT3 ADD3 SUN3 ENC1 CBLC KIF12 ODZ1 ODZ4 SLIT1 GAS6 UCP2 TESC CBR3 SHC3 SHC4 PGK1 EXO1 KIF23 KIF14 CDK1 PRC1 ANLN TUBB KIF15 KIF11 MLH1 FDPS KIF24 LDHA CDK2 GALE CAST DDB2 A1BG MTX1 NDE1 MZB1 GPX1 TPM1 NEU4 RNLS MT2A SDC2 TDO2 SLIT2 CBFB MYT1 ODZ2 GBP2 MT1A PGK2 SGK1 BDH2 PSG9 SDC1 MT1X TPM4 IL1R1 OFD1 KIF27 BRD7 PDCL PLDN NFYB CTBS XRN2 NBR1 GLO1 DCP2 PYGL DSN1 DNA2 TFRC XPO1 NRP2 EDN1 DSC2 GAB1 SOS2 GBP3 ADH5 TMX1 UBR7 CRYZ CALR LIN54 FRG1 PSG4 EAPP UCA1 UBR1 KIF22 PELO CDK6 CALU CBR4 RPN1 TMX2 CDS1 LIN52 GBE1 POP1 GPX3 GPX2 UCK1 BRD8 NRP1 XPOT SACS AOC2 ARTN AOC3 ACPP AGR2 AGR3 COA5 SYBU UACA RCN3 NOB1 PRKX EMX1 SCXB FCAR HCN2 SMA4 DSG3 PSCA GPC1 EGR4 CHD5 STRC MPP4 LRIT3 ANO7 ESPN BRDT GNB2 MAP4 SCAP TMC4 ZXDC IKZF4 TBCD NCLN LMO3 NMT2 CASK TMC7 NKRF GNB1 CHD7 IKZF5 TPBG SPEN CDH3 AARS COBL EME2 POLN SUFU ARSA SOLH ANO8 PREP NEO1 BMP1 PCNT SMA5 CCR6 MPP2 CTSC DOK7 RGS3 EGR3 EGR1 GPC6 MYLK EGR2 DTNA CDH2 FOSB TMC1 DDR2 DOK3 PKM2 HEG1 GPC3 ASPH ASNS EME1 RND3 GPC5 CDC6 MELK SHQ1 GEN1 CDC7 PSPH ENO1 RPSA MAFF PMEL NPNT TIAF1 BMP7 DOK4 APOF SSPN CDH5 CTSH TFPI2 XKRX GLRX PON2 ARG2 OSR1 CNN3 MGLL GNB4 WIPI1 FMN1 CHD9 LRIF1 PHKB DSTN HELQ HEXA MFN2 TPMT RCN2 ENSA GPD2 RRN3 GPN3 ADSS HIAT1 TSFM UGP2 MAP7 TCRA THRB GGT1 GRB7 ANO6 NPTN GDE1 TMC5 GLRB SOD1 GPN1 CTSD CCR8 ANO1 NFIL3 UFM1 USO1 SBDS CAPS CHN2 NENF SARS LMO7 IKZF2 OPN3 HEXB GNB5 MPP6 PERP OSR2 IP6K1 EVI5L AIPL1 PPIL3 PPIL1 CSDA ATAT1 ACCS AGPS QPRT SIVA1 CROT PADI2 PADI3 GATM NIPA1 NIPA2 PDPR CD3G MBD6 TGM2 MEFV SPEG CSAD LAIR1 PTMS SHOX SSPO PXDN MSH5 CMBL RXRA CLIP1 RAI14 MSH4 CERK PLIN1 NADK GGA3 WFS1 PRNP PUM2 NME4 CCNF MPST NPM3 NEIL2 RARA ITPK1 RERE MYH9 BZW2 EIF5B MTG1 GNAZ AMD1 LPIN3 LMNA OPTN MAFB EIF5A GLDC NREP MEST CLIP4 PITX1 NOD2 ANKH ADM2 TYMS TGM3 BCHE NEIL3 KIF4A ORC6 MXD3 POLQ ORC1 GSG2 MTBP EIF4E SMC2 SMC4 DHFR PCNA FMO5 VCAN ECM1 UTRN KYNU RARB SYNC GYG2 CLGN HCG4 SMN2 FSIP1 QPCT FSIP2 SNPH ORC4 NANS DERA GCC2 SMC6 YIPF4 YIPF5 M6PR GCH1 BZW1 CYCS MSH6 PFKM ARSD NOD1 GFER TYW1 LPIN2 MNX1 RHEB TGDS RARS ORC3 MED9 MED8 LIN7B STIP1 PDHX DRG1 PDHB NARS YIPF6 DPM1 OSTC PECR SPOP SWT1 VWA2 VWA1 SIRT7 MYOF OTOR LYZL2 SIRT4 DMC1 PRND RFNG SDHC CCNK AHRR MDC1 NPCR DNM2 MUC2 MND1 RRM2 MDH1 RRM1 MUC1 PNMT CRBN DCXR CHML AMFR RHCE CRCP SDHD URM1 NMD3 GPER KIF5C COLQ DGKB PCIF1 HIF3A GNG4 PGM5 KIFC2 PI4KA INTS3 POGZ SIDT2 GNG7 INTS1 ITFG3 TRIB1 SGSH ILDR1 PPIEL SMG7 INTS9 KIF1C IPO13 GALM TRIB2 KIF3C ESRG GARS KIFC1 AMBP IL1RN APOD KIF2C ASPM PSIP1 CLIC4 BORA GM2A TRIB3 EREG EPOR RHOV BBIP1 WSB2 PGM1 DGKA HOPX COQ3 ISG20 SGCE KIFC3 FNIP2 PGM3 ITFG1 GUSB COQ7 COG3 SIKE1 NAPG TIPRL WBP1 COQ9 CLIC3 INTS2 PGM2 COG6 INTS8 SMG8 CLIC1 COQ2 EMG1 COQ5 AZIN1 ITPR1 AMOT STOM AGRN STAT4 UROS FAIM3 RHOA RORB RORA TAF15 PAICS IMPA1 IMPA2 TAF11 ATP5J TAF13 TAF12 SIAH2 TNNI1 HIPK2 ITGA1 QDPR ZMIZ2 NIPBL SIN3B NCDN ZMIZ1 IGSF9 GRPR UBE2I GRM4 ITGA3 PILRB AIM1L ITPKB SPIN3 RHOB AXIN1 BNIP3 LIMK1 BCAM WNK1 TTLL4 RHBG ITGB5 HMBS MRAS SORD GRM6 CHRD SGCD AXIN2 CHDH SNAI1 MKI67 ITGB2 ITGA6 ITGA5 ITGB6 DDIT4 CAMP SCOC GGCT SLIRP BRIP1 TMPO MARS LRIG3 GANC SNCG HNMT SCGN BNIP2 TTLL7 MIS12 CCNC VIPR1 ITGA9 IGSF3 MCEE IGSF8 HINT2 AKTIP PDIA4 GAMT RBMX OXSM IARS2 ISCA2 HINT1 BNIP1 APOO BNIPL RIPK4 RERG ITGB8 CRYM MAOB FANCI RORC ATP5L BICD1 KLF13 VSIG1 MILR1 NWD1 IGF1R WDR6 BICD2 LTBP3 RHOU MDM4 CDON COMT AIFM1 LTBP4 KLF16 IGF2R STIM1 KLF10 DIP2C MECR TIMP3 TIMP2 COCH BIRC5 RIBC2 MCM6 TRAIP UGDH MCM8 MCM5 MCM3 MCM4 MCM2 MCM7 SGCG RHOH TGFBI CRIP1 TIAM1 MYCN VSIG2 LTBP1 WDR5 POMP NGDN MDM1 CISD2 COPG CRIP2 TIGD4 BPGM IDH3B IDH3A WARS GGCX NGRN GMPS RMRP NRIP3 NFAT5 ATAD2 ATAD5 ATHL1 ATF6B NAT10 NAT8L TAF1A TAF9B MLXIP EP400 GRIK4 ZFP42 SALL4 MLLT1 ZFP36 TRIM8 ORAI1 RHOG OSMR ITGAX P775P LLGL2 C9orf3 UGCG SP100 ZFP41 ITM2C ORAI2 EP300 CHID1 FSTL4 LZTS2 PTGIS MLLT6 LLGL1 GRIP1 VWDE GRIK3 GINS3 GINS1 GINS4 GINS2 S100P C8orf4 FSTL1 FSTL5 MEIS1 MSMB RHOQ P704P ZFP14 C1orf9 HBXIP GNAI3 DIRC3 MLLT3 CNIH3 CRIPT BLZF1 AIMP1 ITGAE CNIH4 C4orf3 DIRC2 GRIK2 CLYBL TFAP4 LYPD3 FLRT3 GATA3 GATA2 GATA6 KAT2A PALLD ATP7B FRAT1 ATP9A FOXJ2 KAT6B TAF1C SATB2 PSAT1 ATP5E LOXL1 LOXL4 ETAA1 LPAR6 LPAR3 ATG14 STAP2 ATG12 LRP12 LRP5L TCF25 TEX11 INHBB BCL9L SYT15 TTC28 SYTL4 MIER2 SF3A1 ZZEF1 ADAT1 STK38 LRP10 SF3A2 FBXL8 ZNF91 TCF20 ZNF74 WIPF2 FBXL4 TTC18 A2LD1 RPL13 HYAL3 ADAT2 SF3B4 FURIN RPL28 STK10 SCRIB VAC14 MINK1 TEX19 ZNF69 PLCL1 PRINS EMID2 RIMS4 SYTL2 ICAM1 CENPI RPL31 RPL39 RPL36 PRIM2 PRIM1 GMNN TCF19 KLHL2 STX19 TEX15 TTC14 PNISR ZNF45 ZNF12 ZNF41 MIER3 MIER1 MIEN1 FBXL3 TTC26 RPL41 RPL21 ZNF24 TTC35 RPL26 SYT12 RIMS1 ZNF22 FBXL5 RIMS3 RIMS2 TTC33 RPL34 STK39 RPL35 RPL27 RPL32 RPL37 H2AFJ ZNF79 CIRBP STAC2 ACTA1 ACTA2 ACAT2 FLOT1 MAGI1 KRT23 KRT81 KRT15 KRT13 KRT80 KRT18 KRT10 mir − 34 FABP6 FABP5 SSFA2 TVAS5 BCAT2 NEAT1 ATP5D ATXN2 ATP5H ATXN3 PHF16 CD209 CRIM1 CD163 IMPG2 SULF2 ASB11 CCL16 TCL1B AP2A2 ASB13 CCL22 SPPL3 ASXL1 NISCH YPEL3 LASP1 PEX26 TPD52 LUC7L TULP4 CD276 WWP2 PELP1 AFMID VPS53 FSD1L LCN12 ALPK3 ESPL1 PLEK2 PHF19 HMMR CTSL2 FDFT1 AP2S1 TEKT5 SYPL1 ECT2L TPST1 ALPK1 CCL28 VEZF1 FBLN2 EPAS1 SAP18 VPS41 BBS10 SEP15 H3F3B VPS45 ZBTB6 UTP23 VPS35 UTP15 FGF13 TTBK2 PEX13 SULF1 ALAS1 VPS25 PHF17 ALG14 PEX12 STT3A ALG10 HIBCH ATCAY P2RY6 SYAP1 TACC2 CDYL2 PACS1 TAOK3 PACS2 TACC3 TACC1 STAG2 PLAC8 ACPL2 FADS3 ACSL6 FARP1 ACSL4 FADS2 ACSL3 ACSL1 PLAC2 PVALB ZMAT3 PAPLN PARK2 ATG9B TAPBP PANX2 HSPA5 ATG9A ATG2A TANC2 PANK1 PARP4 HSPA2 NPAS2 PAPD5 ZMAT2 NTAN1 HSPA9 PARP2 PARK7 ABTB1 USP35 ZFHX3 PVRL3 KCNJ5 NYAP2 MYRIP ZC3H4 PPEF2 SNX29 SCIMP P2RX5 IFITM3 PTPRJ SOX12 TERF2 RNF26 RNF44 SNX22 TLCD2 SOX13 RLTPR PVRL1 PLBD2 RAB42 ABTB2 RAB30 LSM10 PTBP1 EGFL7 GNL3L SNX30 USP20 SH2B1 USP21 BAZ2A RNF43 USP54 TESK1 ZG16B LPPR2 SNX25 USP14 SEC63 CEP68 FOSL2 SAFB2 RNF24 AUTS2 NR2F6 RAB26 ARL4A LRFN4 LARP1 USP36 LRFN3 ESYT1 ZC3H3 ZFHX2 TTYH3 TSHZ3 TRAF7 NAA60 S1PR2 WWC1 TESK2 TELO2 TTC7B PVRL2 USP10 SEPT3 STRF6 FKBP5 NR2F1 LRFN1 DPP10 RAB32 S1PR3 SH2B3 TTYH2 IFITM1 HELLS SEH1L CEP55 SPC24 PPEF1 ULBP2 H2AFZ RPLP2 ASF1B SPC25 RPS18 P4HA1 U6atac RPLP0 RPS24 RPS12 CEP78 ZWINT RPS19 RPS15 BLVRB BSCL2 RPS29 RAB20 SEC13 RNF39 RAB34 TFCP2 PLBD1 TRAF1 FKBP2 MDFIC RPL2B TCTE3 USP51 S1PR5 SNX10 HYLS1 ASCL4 P2RX2 ERP27 RAB19 RAB31 BFSP1 LRFN2 PLCE1 TSTD2 LARP6 SNX13 SRP14 KCNJ3 TSHZ2 PVRL4 ASTE1 RNF13 DPY30 CEP44 NR2F2 PTBP3 SNX14 RNF11 CSTF1 CSTF2 RNF25 SCYL3 USP12 ASF1A RPS11 CEP57 NAA38 SEPT7 HBS1L USP37 CSE1L SPEF1 NAA40 RAB15 PLCB3 USP25 BLVRA BAZ1A CRYL1 NEK10 LRP1B N4BP2 FKBP8 RNF40 NOL11 SNX21 RAB23 AKR7L CLCF1 SNX16 SEC62 USP16 PHTF1 TRAF5 SRP54 RPP14 P4HA2 TRPT1 JAGN1 NTHL1 SNX11 PHTF2 NAA20 RPS20 RPS25 RPS26 CEP97 RPS14 CAB39 ZNFX1 PUS10 SNX33 SPEF2 N4BP3 P2RX4 DOT1L COTL1 ATOH8 ACSF3 SPAG5 ACSF2 SPAG4 LACC1 MAGIX SPAG1 AGBL4 LAGE3 VASH1 LYRM2 FANCL HLA − A LYRM1 DACT3 BUD31 LRRK2 PRR19 ASTN2 TUBB3 PLCD3 EFR3B CBLN2 DDX10 GSTT2 ZBED1 SCAF4 CHTF8 CPT1A FOXP4 FOXK1 PRR12 NPRL3 ZNRF1 PRR22 NR2E3 PPM1J NXNL2 FNBP4 LETM1 UCKL1 PRPF8 EFNB1 CASZ1 EFNA1 RFTN1 TRPV1 MLST8 DMWD LMTK3 HSDL1 TSEN2 CDK19 NBPF1 CTBP2 PPTC7 CPSF1 LRRK1 PTPN3 RNFT2 PRR24 DHX37 RRP12 NHSL2 LPHN1 SCAF1 BTBD2 ZNRF3 NFKB2 DHX33 PRR14 DDX51 DHX34 CDK10 PNPT1 ZBED4 TEAD2 ETNK2 TEAD1 HHLA1 SSTR2 CYR61 TUBB6 TGFB1 NR4A1 SFXN3 DGAT2 PDZK1 DDX60 THBS1 PRR11 TUBB1 DDX11 DDX21 TGFB3 GDF15 RAD51 UBE2T PTTG1 H2AFX GHITM DCLK1 SFXN1 RAD18 VLDLR TNNT1 TCEA3 DDX25 ARL4D LPHN2 AJUBA MPZL3 GFPT1 PLCH1 CDK14 DLEU2 COX17 ARL4C DHX29 RNFT1 BPNT1 GTF2B TTC9C THAP9 ZBED5 H2AFV UFSP2 DDX58 HSDL2 NUP93 PRPF4 OSTF1 RAD21 CPSF2 DERL2 CEPT1 MAT2B SRSF7 CCBL2 SRSF1 SRSF3 DUS4L NUP37 TCEB3 SERF2 BCL7C NUP43 DHX40 CRLS1 ZBED6 BTBD6 NUP54 GSTZ1 EEF1D THAP6 CTPS2 DDX23 PRPF3 EFNB2 ETNK1 DHX15 LETM2 NUPL1 THAP5 THAP3 RRP36 ETHE1 TCTN2 PARD3 TUBE1 PHPT1 THAP2 DUS2L SRSF9 PHF5A MPZL2 EFNA3 TRAK1 EFNA5 EFNB3 ASAP1 APBA2 SSBP3 VPS4A OBSL1 PEAK1 AKAP1 APBB2 SPRY2 SPRY4 POLE2 SASS6 POLA2 SKAP2 PEG10 BOLA3 AKAP5 VPS4B AKAP9 GALK1 AKAP6 G3BP2 POLE4 POLE3 SEPP1 SPSB1 SSBP2 ADAP1 PAQR7 DAPK1 PAQR4 TOP2A PAQR8 TOP2B TRUB1 RUFY3 ACAA1 BACE1 QTRT1 PRRT3 WWOX ACTN1 ACTC1 DYRK4 HPRT1 RTKN2 HLA − C ACSS3 LRTM2 ACAP2 ACTR6 ACSS1 ACYP2 ACTR3 OVOL1 MFAP5 ERO1L MFAP3 MFAP1 PPARA MATN2 BUB1B ACOT9 ACOT7 ACOT4 ACOT2 VDAC3 MNAT1 ATPIF1 VDAC1 RGS11 FOXN3 PCBP1 PCBP2 PDE3B NOP14 LRRC2 EFHD2 GRB14 HEBP1 PMP22 PREX1 CDH23 CABP4 PDZD4 PDE6A LAMA1 FOXC1 OR2A7 GPR83 NDST3 CHST1 ZFPM1 GPR56 CHST6 NLRC3 GPR65 ABCB5 PDE6B ABCA3 GPR82 UPK1B LENG8 CYTH3 THSD1 MYLK3 FGFR4 FGFR2 FHAD1 THSD4 GPR68 PHRF1 ABCB8 CLCN7 UBAP2 CTDP1 CYTH1 KLRG2 PDPK1 PDE8B NDST1 RAB7A GPR21 SPNS1 HSPB1 LRCH4 ABCA7 KANK1 TPCN1 SPHK2 SEPN1 REPS2 UBE2B TPCN2 CDC16 PLCG1 ALMS1 LRCH1 DCAF4 BARX2 CPEB2 GPR52 MCF2L PDZD8 CLCN6 SYNE1 NDST2 ERBB3 CSPP1 FCRLB SCFD2 POLD1 MZT2B TRNP1 NOP16 RGS12 PIWIL2 YLPM1 EPHB3 LRRN2 GPR35 SYVN1 CLDN3 PLOD3 ABCA2 GNA12 PTPRF LAMA5 LMX1B TCOF1 RASA4 CPT1C DCAF7 FBXO2 GPR77 MYBL1 CLCN5 FGFR1 ABCA9 EEPD1 BRSK2 CTXN1 EPHA6 LRRC3 PHKA1 ENPP5 ENPP4 GRB10 KCNF1 PDZD7 OXCT2 HSPB8 RASA3 SESN3 EPHA8 ARNT2 LRRN1 RAB3B TUSC3 CPEB1 NR1D1 SRPK1 PLOD2 INSIG1 CDC20 MYBL2 RGS16 KPNA2 MS4A7 ANXA1 CKS1B PDE5A FUCA2 UBE2S NDC80 GTSE1 CDC45 RBBP8 HSBP1 CKAP2 RNU11 CETN3 RPS3A PRPS2 KNTC1 FMNL3 FBXO5 CYB5A PRPS1 SRPK3 AZGP1 UCHL1 EPHA1 EGLN3 ARAP1 CDH17 PDZD2 GPR19 ENPP3 BEAN1 RGS10 CASP4 GPR39 GULP1 LAMB1 ENPP1 UPK1A OXCT1 SYDE1 BANK1 EPHA7 NOP10 PTH2R PDE4B EPHA4 SPNS2 GNA14 EPHA2 ABCA4 CDH10 STRN3 REEP1 PSEN1 NR1D2 ABCA5 PREPL GPR98 UBE4A CASP6 SENP1 VMA21 AP1AR FBXO8 TM2D3 EXPH5 UNC50 GLRX2 CETN2 RAB2B SURF4 PFDN1 FOXN2 ZCRB1 KCTD6 NR3C1 SENP7 SREK1 PDE7A CDC5L SYCP2 SESN1 PPP6C KPNA5 CDC26 PLRG1 FBXO4 POLD3 FBXO3 ZUFSP PFDN4 RDH11 CCT6A CYB5B CDC27 APH1B RDH10 PQLC3 CFLAR C2CD2 BCAS1 LAMA3 ABCA1 RAB9A FGFR3 GSTK1 TM2D1 CES4A PLOD1 NIF3L1 CCT6B PFDN5 VTCN1 LCMT2 RAB8B ARAP2 RAB1A SAR1B CASP3 HEBP2 MYL6B FRRS1 CPEB4 BCAS2 ANXA2 SPHK1 BCAS4 SENP8 FUCA1 SCFD1 AHSA1 RBBP9 CDC42 ABCE1 SAR1A RAP2B SURF1 CASP7 ESRP1 ANXA5 RC3H1 ASAH1 ERAP1 PRSS8 ERBB4 FOXD2 RAB6A PCSK6 RGS22 INSIG2 HAUS6 HAUS2 HAUS4 RTDR1 NUAK2 CRTC1 POTEF ACOX3 CRTC2 STON2 TONSL RTN4R ACOX2 DARS2 TOR1B OTUB1 VOPP1 AGAP5 AGAP9 AGAP3 ACTG1 AGAP1 NFAM1 MYO19 ACTG2 AGAP6 VAMP2 AGAP8 AGAP4 VEGFA FANCA FANCB BACH2 ACER1 AGSK1 ACBD7 VAMP5 FAM5C EBAG9 BACH1 AGAP2 VAMP8 ACBD3 DAGLA HDAC5 HDAC4 APOA1 HDAC9 CNTLN FOXO1 CNTN4 HCFC1 TMTC1 CNTD2 TRMT1 FHDC1 MGAT5 HTR1D PAMR1 MGAT3 PARM1 CENPA NUDT1 HJURP UHRF1 NUDT6 ENOX1 CNTD1 HOXB2 COX8A HOXB3 CNTRL HOXA3 PALMD GABPA HCFC2 TMTC2 CCNT1 HOXA5 NIPAL3 CNTN1 TMTC3 NUDT9 NIPAL2 ZNRD1 COX5B DCTN6 DCTN5 COX7B MGAT2 MTFR1 FOXO4 MEF2B KIF26B PCGF5 CPNE7 NPHS1 GFRAL KIF26A ABCC8 CERS1 EIF5A2 AS3MT ABCC5 KCNE4 MYD88 PXDC1 DUSP8 PGAP2 ADCY9 MYH14 MED25 PDE4C VENTX CCBE1 ASMTL NRBP2 KIF18B DUSP2 KCNB1 NXPH3 RBM20 NOC4L MED24 MED12 ARVCF MED26 RREB1 ABHD4 ARRB1 CAPN1 RPPH1 ABHD2 ADCK1 BCAR1 MED16 MTSS1 BCAR3 GSK3A PPDPF TNPO2 EIF4E3 ABCC1 RNPS1 SUSD4 SCN8A DUSP1 RBM33 MAST2 SNED1 PRKD1 CLSPN QPCTL MRPL4 CAPZB CERS4 LAMC1 PCBD2 MAST1 PCGF2 PTPRS RBM38 NPHP4 DUSP7 ADCK3 NOLC1 MED15 MAST3 KCNK3 ABCC4 ADCY3 NLGN2 PPRC1 KCNK5 MYH15 RBM24 OLFM1 GP1BA ADCY2 SUSD3 PGAP1 LDOC1 IL17RB SCML1 GFRA1 PYCR1 DUSP6 ADCY5 DUSP4 SGOL1 UBE2C KIF20A KIF18A CASC5 RASD1 BRCA1 LMNB1 CHEK1 BRCA2 CSRP1 KIF20B PSRC1 KCNS3 CALM2 TPRKB GLUD1 CENPL BARD1 MASTL LMNB2 CHEK2 PRDX4 TUBG1 SERHL GSK3B GRHL3 ABHD8 BEND6 ABCC6 CSRP2 NXPH1 CGNL1 PCBD1 IL27RA PKHD1 TPI1P2 PTPRE COPZ2 GRHL1 IL20RA PDE4D SRBD1 PRKD3 SGOL2 CAPN8 PRRX2 SGPP1 MBNL2 L32131 SCN1B RSBN1 EIF2S3 LCORL CASC4 NPHP3 BECN1 LEMD3 RBM18 CAPN7 ARPC3 CPNE3 HARS2 LEMD2 UBE2N HSPH1 DMXL2 JKAMP EIF4A2 TSNAX SPOPL ASCC3 PRDX3 EIF2S1 RBM23 PGAP3 CAPN9 RRBP1 CPNE2 CAPN2 BSPRY MRPL1 PCGF1 MBNL1 CNPY2 GLUD2 FEM1B CASD1 NUBPL EIF2S2 MED31 PRDX5 NUBP1 BAHD1 ABCC3 SUSD1 MED23 COPZ1 OVCA2 UEVLD MRPL3 PCGF6 FOPNL LMAN1 EIF2B3 CMYA5 KIF13A ABCD4 ABHD3 PTPRU DAND5 RHOT2 OTUD1 OTUD5 RUNX1 CNOT3 GSTO2 CNOT8 CNOT7 CNOT1 RUSC2 MACF1 ACCN3 NACC2 ACCN2 OVGP1 CHAC1 AGGF1 PDGFA ALDOA PPARG PAIP2B AMY2B MUC19 DNAH5 HOXC6 BMP8A GLOD5 CARD6 HOXC5 PDCD4 PODXL DUOX1 PTCRA BHMT2 GALNS UNC5B IL17RD HOXC9 REXO1 INPP5J TRPM3 SPON2 TRPM6 RNPC3 MUC20 MAP1S KCNC3 HSPG2 PDDC1 DCHS1 SOCS2 BMP2K HOXC4 ACACB MAPK7 EIF2C1 PI4K2A NPDC1 SBNO2 DEGS2 RPRD2 CECR2 MEF2D TRPM4 MFSD3 AHDC1 NUCB1 CDCP1 PRKG1 ERCC6 LARGE MTUS1 FREM2 ACACA DNAH1 TRRAP KCNH3 QSOX1 CECR6 TRABD NPTXR RHPN1 SOCS3 DECR2 HERC4 UBE2O EHMT2 STMN3 PTGES GREB1 DHRS2 KCNC1 CASQ1 BMP8B KCNH1 NEURL HOXC8 THRSP NUPR1 MEX3A CXCR4 CDCA3 FOXM1 CDCA8 CCNB1 DEGS1 SHMT2 CCNA2 CDCA2 CENPF CDKN3 CCNB2 GCNT3 XRCC2 TRIP13 CDCA5 CDCA7 ESCO2 STMN1 DSCC1 COX6C CCNE2 POC1A SHMT1 FHOD1 PSMA6 SNRPF COX7C C1QBP PXMP4 UNC5A NCALD DSCR6 TMED3 NSUN7 LMCD1 HSH2D RFESD MAPK6 PSMB8 MEF2C NEDD4 KCND2 MUCL1 GNG12 KCNU1 PSMB7 TRIP11 PRMT7 DSCR3 PNRC2 TMED5 NEDD9 COPB2 PCDH9 ABCG1 PRRC1 PSME1 TRAM1 FCHO2 PPM1B ESCO1 TMED2 COPB1 PSMA2 EXOC5 PSMA1 DHRS7 THOC7 SLMO2 NEDD1 XRCC5 PTPRH RHPN2 DHRS3 CEBPB CXCR7 VEGFB HERC6 PPM1K MUC16 ARCN1 SHFM1 OSCP1 PGBD2 HCAR1 PSMA4 MAP1B KRCC1 GCFC2 NCEH1 PI4K2B PXMP2 PSMB5 SCRN1 PSMA5 PSMB3 XRCC6 PSMB4 PSMB1 XRCC4 MFSD1 COPS8 TMED1 COPS4 PSMA7 COPS3 PSMB6 SOCS6 PSME2 DEDD2 EIF2C4 HERC3 WASF1 WASF2 AURKB AURKA TOLLIP ADAM8 TROAP MYO9B HAGHL MYO3B OTOGL ACSM1 ACSM3 MAGT1 MYO5A AAGAB NCOA6 NCOA1 MLANA KDM4B MEGF6 RAMP1 X05128 ARGFX DCDC5 GDPD5 NPIPL3 SMAD9 CEBPD MYH7B PRKCA BPIFB1 MEGF8 SRCAP CDHR4 KHSRP GFOD1 KDM6B DOC2A FCGBP PRKCE SHOC2 SAMD1 TTLL11 FBXW4 SMAD7 TMCC1 MMP15 GSTM3 KCNG1 TMPPE PIK3R2 SAMD8 SMYD3 UNC5C DDHD1 MARS2 MAML1 DNHD1 PDGFB MMP17 MMP25 DNMT1 FRMD8 SNHG3 NPIPL1 DOCK6 MARK4 SIGIRR CCDC6 MEGF9 DOCK5 EIF4G1 SMPD4 DOCK3 AMPD2 FRMD3 PNMA1 OPRD1 PTPRG MTND5 BPIFB2 KCNG3 MGST3 DHCR7 PPFIA4 CENPK CENPE CENPP SNRPE MARK1 ERMP1 SNRPB PSMD5 ESRRA SMPD3 MMEL1 GDPD1 RCCD1 ALDOC MSRB3 MGST2 SQRDL NEGR1 FBXW7 CABYR MGST1 SMAD4 MTHFS SMAD1 AMPD3 CMPK1 OSTM1 EIF4G2 CREG1 PDIK1L PSMD4 TMCC3 RGPD6 NARG2 RBMX2 FRMD6 PSMC6 BNIP3L RECQL EIF1AX DOCK4 PRRG1 CCPG1 PSMC2 PRRG4 MCFD2 EDEM3 PSMC3 PSMD8 PSMC1 PSMD6 CENPV RBM8A PSMD1 NTPCR SRPRB MRPS7 RNGTT GDPD3 PIK3R3 DCDC2 MARK2 SMAD6 SMAD3 RPTOR MERTK RGAG4 MACC1 ROBO1 ROBO2 TIPARP WNT9B TMOD2 FBLIM1 WDR31 WDR81 WDR24 SGSM1 MOB3A WDFY2 WNT9A CHERP NUMBL NCOR2 SRRM2 HMOX1 CARM1 HEXDC SGSM2 CYFIP2 CADM2 WDR27 WDR37 PRKCD SGMS1 ADPGK PRKDC MARC1 WNT7B SYMPK GPSM1 BAIAP3 GMEB1 WDR59 MTFMT WDR90 DRAM1 SYNPO GOLM1 WDR62 DNALI1 CSMD3 HHIPL2 MLF1IP GPSM2 NCAPH CENPN ARMC3 DTYMK MCM10 CENPH PGAM4 WDR76 FIGNL1 PGAM1 PSMG3 SEPW1 WDR34 TMEM9 PPCDC PRDM5 ADPRH GABRP PRKCH TMCO4 WDR67 PDGFD CCNG2 ZMYM5 KHNYN ARFIP1 MKRN1 ZMYM2 LZTFL1 MPDU1 PSMG2 WDFY1 VSIG10 ZNHIT3 GSTCD TMCO1 WDR54 UHMK1 ARMC1 SSX2IP WDR73 SCMH1 TMCO3 SGMS2 NCAM2 PIH1D2 CADM1 PDGFC MOB1A FILIP1L TMOD3 CHMP5 WDR61 SNRPC DRAM2 TRIAP1 PBXIP1 SPINT1 POTEM DAB2IP MYO1G CORO6 STAT5B ATPAF1 ATP5F1 C5orf64 RAB3IP C18orf1 TRIM66 C9orf72 C9orf89 WFDC8 C4orf26 C5orf45 CNNM3 C3orf62 C5orf42 HIF1AN HIVEP3 HMGA2 NMUR1 TRIM58 PRKCG MLXIPL C7orf46 C2orf83 CNNM2 C1orf86 C4orf44 ISYNA1 C10orf2 SIRPB2 PIEZO1 C7orf43 INPP5K TRIM62 WDTC1 PIEZO2 C4orf48 C2orf18 MON1B C8orf55 C5orf30 NFKBIZ TRIM74 C3orf58 C9orf93 C7orf41 MLLT10 C12orf5 RMND1 IGFBP4 HMGA1 SHISA5 IGFBP2 GABRD IGFBP5 C15orf2 SPIRE2 MDGA1 PIK3CA MCCC2 C9orf85 TRIM25 TRIM47 C1orf21 TRIM36 ABLIM2 IGSF9B ABLIM3 CNNM1 C3orf80 C3orf67 SEL1L3 OBSCN OGDHL C9orf47 C2orf66 MTMR2 S100A6 S100A1 ERRFI1 NCAPG ESRRG HMGB2 C5orf34 HMGB3 MIS18A SS18L2 SNRPG C4orf21 TRIM59 CENPQ CENPO GAPDH C2orf72 CMTM1 USMG5 C2orf54 C8orf85 MDGA2 MLLT11 C1orf56 MOCS1 GCHFR C8orf75 C11orf1 C9orf64 C7orf58 C3orf14 C6orf99 S100A9 SPINK8 MCCD1 TRIM46 C4orf38 SHISA2 IFNAR1 C1orf63 C6orf47 C7orf44 ZFP112 ZFP106 TRIM45 NDFIP2 TRIM23 C3orf38 INPP4B C1orf27 PPWD1 CMTM6 C2orf47 C8orf76 DTWD1 C7orf23 C5orf44 C4orf33 C5orf43 UBE2W C1orf96 TWSG1 C4orf19 C8orf77 C2orf68 ABLIM1 C6orf72 C4orf32 EPCAM C3orf52 C2orf43 MYCBP C3orf23 C8orf40 SUMO3 C2orf44 C9orf82 C4orf39 HMGB1 C12orf4 C5orf15 SUMO2 C20orf3 C5orf54 TRIM69 C9orf91 C8orf45 C2orf28 GMPR2 C8orf83 C5orf35 C3orf75 SUMO1 C14orf2 C8orf59 C4orf43 C5orf32 C6orf57 C4orf27 C5orf41 CMTM4 CITED4 C3orf70 C6orf97 SHISA9 HILPDA TOMM7 TOMM6 TOMM5 BRWD3 RRAGC BRWD1 PRODH CRIPAK ELAVL2 ATP1B4 ZNF215 ZNF618 ATP8B2 FBXL16 ZNF814 ZNF865 FBXL13 ZNF677 KLHL21 ARID3B ZNF208 ZNF598 ZNF296 ZNF580 MYADM ZNF703 TCF7L1 ZNF662 PIK3CD ZNF689 ZNF835 ZNF414 ZNF827 ZNF142 ZNF628 ZNF773 MICAL3 NKAIN1 KLHL36 PITPNA WHSC2 HUWE1 ZNF777 ZNF517 ZNF646 ZNF552 ARID5A ZNF192 ZNF496 ZNF213 FBXL18 ZNF358 ZNF579 ZNF761 ZNF737 PITPNB ZNF439 ATP1B3 ZNF749 ZNF350 ZNF708 ATP1A1 ZNF248 ZNF740 ZNF431 ZNF676 DNMBP ZNF726 ZNF326 ZNF813 ARID1B ZNF516 ZNF384 ZNF850 DIAPH1 ZNF556 ZNF503 ZNF276 ZNF577 SEZ6L2 ZNF613 ZNF467 ARID1A ZNF219 MGRN1 KLHL38 ZNF341 ZNF490 ZNF527 ZNF273 ZNF335 ZNF692 MORC2 ZNF649 ZNF165 ZNF239 ZNF238 ZNF486 MICAL2 MICAL1 ZNF587 PDLIM7 PDLIM1 ZNF727 PDLIM3 DIRAS2 HMGN5 LRRIQ1 DIAPH2 ATP1A4 ATP1A2 ATP11A ZNF714 ZNF730 ZNF695 DIAPH3 PHGDH RPL39L CRISP3 CENPM ZNF367 ZNF257 ZNF519 ZNF732 ZNF492 ATP8A2 PDLIM4 INPP5D HMGCL ZNF655 ZNF396 ZNF512 ENDOU HECW2 PDLIM5 ATP2B4 MORN2 ZNF678 ZNF608 ZNF750 KLHL13 ZNF175 ZNF266 ZNF271 ZNF624 ZNF287 ZNF638 ZNF226 ZNF187 ZNF879 MORC3 ZNF224 ZNF644 ZNF302 ZNF181 ZNF227 ZNF304 ZNF330 ZNF322 ATP11B ZNF277 ZNF572 WDR5B ZNF134 ZNF155 ZNF658 ZNF260 ZNF133 ZNF337 PET117 ZNF136 ZNF184 ZNF484 ZNF844 ZNF323 ZNF197 ZNF770 ZNF790 ZNF449 STK38L ZNF606 ZNF654 ARID4B ZNF483 SF3B14 ZNF410 ZNF189 ZNF627 ZNF404 ZNF267 ZNF180 MIR23B ZNF747 HOMEZ ZNF830 CMAHP ZNF532 NFKBIA ATP8A1 ZNF223 ZNF143 HMGN1 ZNF510 ZNF200 ZNF230 ZNF232 ZNF630 GORAB ZNF268 ZNF222 MORC4 ZNF295 ZNF221 GMPPB HMGN3 ZNF684 ATP5A1 UQCRB HMGN2 ATP2A3 KLHL24 ZNF596 ZNF853 IMPACT LY6G5B DYNLT3 ACTL10 TATDN2 ATAD3A CROCC ROMO1 ATAD2B TATDN1 TAS2R5 ATXN2L ATP2C2 LPCAT1 ATP10D SFTPA2 NRCAM CIRH1A ZBTB20 GIGYF1 FER1L4 ZBTB46 IQSEC1 ZBTB42 IQSEC2 ZBTB45 TIMM23 ERICH1 SAP130 FER1L6 TICAM1 MMRN2 LIMCH1 WDHD1 MSMO1 UQCRH CCL3L1 UBE2J1 LHFPL2 GIGYF2 IFNGR1 ZBTB41 FTSJD1 ZBTB11 TIMM22 ZBTB26 GCOM1 TWIST1 DIXDC1 DGUOK DIABLO IFNGR2 TIMM10 PIK3IP1 AGPAT2 NFATC1 UAP1L1 NFATC4 TFAP2A LY6G6C TFAP2B ATAD3C PARP11 GPRIN1 RNF207 RNF125 GATSL2 ZC3H18 KCNJ11 MINPP1 RNF223 RNF167 TTC3P1 RNF115 ZNF37A IMMP2L RNF187 RNF208 RNF166 GATSL1 RNF157 CENPW KCNJ18 HSPA14 UQCRQ ATP5G3 HEXIM2 PARP12 PARP10 IFITM10 RNF181 ATRNL1 RNF141 OCIAD1 RNF114 YEATS4 GTF2F2 RNF138 RNF149 TTC39A TTC30B SFT2D3 TTC39B RNF111 TTC30A HIGD1A ERGIC2 YJEFN3 NT5C3L RNF139 SFT2D1 HSPA4L HSPA13 RIMKLB SLC9B2 SLC4A1 SLC2A1 A4GALT AKT1S1 CEP104 SLC9A1 SEPT10 CEP164 UBE2L6 RPL27A FKBP11 SLC4A3 SLC5A5 SLC6A9 SLC7A2 ELOVL7 SLC6A6 PTP4A3 SLC7A1 SEPT11 TSPYL5 IQGAP2 ELOVL2 SLC7A5 IQGAP3 ELOVL5 STK17B RPL35A FLVCR2 CEP152 RPL18A PCP4L1 SLC1A5 SLC9A5 RPS17L SLC3A2 GXYLT2 ELOVL6 IQGAP1 SLC1A4 TSPYL4 ALS2CL SLC4A7 STYXL1 SLC9A7 ARL13B IER3IP1 EEF1A2 SLC6A4 RPS27L CEP120 YAE1D1 BRP44L SLC9A2 RPL23A RPL37A EEF1B2 SLC1A2 STK17A SPAG17 KRT42P PLAGL2 NKX3 − 1 ALYREF ACTL6A FAM83F GPKOW SPTAN1 MALAT1 PAPSS2 TSPAN4 TSPAN5 ATPBD4 TSPAN9 TSPAN1 HS6ST1 FBRSL1 BTBD16 ZBTB8A ZBTB7A TPTEP1 NLRP12 SH3D21 PTPN23 STK32C PTPN18 PTPN14 PRR14L VPS37B SPPL2B NBPF11 ZFYVE1 GAS2L1 TP53I11 GLTPD1 NUP214 VPS26A HS6ST3 CHTF18 TIMM8A CYP4F8 CXCL14 SPTLC3 BTBD11 RAD54L LGALS1 TMBIM1 GAS2L3 PFKFB4 INCENP TUBAL3 LGALS8 BTN3A1 BTN2A2 TCEAL8 PTPN21 H2AFY2 LGALSL THAP10 CYP4Z1 PHF21B TMBIM4 SUPT7L ZNF75D VPS33B TCEAL1 SPPL2A SPTLC1 PEX11B THAP11 BCLAF1 HS2ST1 OSGIN2 CCP110 PTPN11 NUP205 NUP155 PEX11A PFKFB3 GTF2A2 SRSF10 ALG10B HIBADH PRR15L TSEN15 TIMM8B LGALS3 TOB2P1 FAM21B FAM40B ZFAND4 FAM78A − 1 ERVV FAM53B FAM43A FAM13B FAM69B FAM83B FAM65A FAM21A FAM63A FAM18A − 2 ERVV FAM84A FAM64A FAM72B FAM72A FAM86A FAM65B FAM83A FAM13A FAM69A ZFAND1 FAM60A FAM46B FAM82B FAM96A FAM49B ZFAND6 FAM84B DNAJA4 ACOT13 DNAJA2 DNAJA3 PABPC4 PABPC1 SIPA1L3 STARD9 PABPC3 SIPA1L2 ZWILCH HEATR4 MECOM ATF7IP2 STARD5 STXBP1 MS4A10 SH3TC2 LRRC69 FBXO31 CYP2E1 DCAF10 EPHA10 SEC31B TBC1D3 LRRC27 LARP4B UTP14C MCF2L2 CASP10 KCTD13 CYP1A2 FBXO40 DFNB31 TSGA13 ANKLE1 UBAP2L PKD1P1 C2CD2L PNPLA7 RAB11B PLXNA1 PLXNA3 GPR135 GEMIN4 CLSTN1 NBEAL2 FBXO32 PLXNB1 TSGA10 GEMIN8 LRRC56 LEPRE1 PLXNB2 PPP6R1 FBXO46 METTL8 FKBP1A DNAJB6 SH3BP5 CHST15 TBC1D9 ABCA10 CYP4V2 SH3BP4 ANP32B UBE2E2 FBXO44 PLXNB3 COL7A1 PNPLA6 FGFRL1 CHST12 CHST10 HTR7P1 LRRC24 RABL2B TBC1D2 FBXO41 DPYSL5 GPR153 CYP8B1 PRSS23 COL4A5 PNPLA3 ARPP21 STEAP3 PLXNA4 CLSTN2 ANP32E TM4SF1 STEAP1 MS4A14 LRRC31 CKAP2L GPR180 TM7SF3 RPS15A RPS27A METTL4 AKR1A1 PRSS27 CYP1B1 GPR110 TBC1D1 FBXO43 PDE11A STEAP4 ABCA12 GPR160 RAB27B RAB11A RNF19A CLDN12 SH3TC1 SEC23A TBKBP1 AKR7A3 COL9A3 GPR126 GPR162 CLDN23 DNAJB4 CYP4X1 PNPLA8 VPS13C PPP6R3 GTF3C5 METTL9 RAB33B PPP4R1 TM9SF1 LSM14B TM9SF3 DNAJB1 ANP32A SEC11A FBXO28 MYL12B PPP1R2 SEC23B UBE2V2 RAB22A FBXO22 GEMIN2 LRRC40 COL9A2 SH3RF1 DCAF11 COL5A2 ZBTB7C GPR137 MYL12A ARL2BP LRRC48 SNAP23 CHST14 ARPP19 TBC1D8 KANSL2 TM9SF2 GTF3C3 PDZD11 POLR2L METTL6 GTF2H5 SEC61B SEC24A MPV17L SEC22B BCAP31 GEMIN6 FBXO16 GOLT1B GPR146 ARNTL2 CYP1A1 PHLDA1 PHLDA2 PHLDA3 CTAGE5 RUVBL1 ACAD10 LYSMD3 LYSMD1 FAM41C FAM83H FAM21C FAM13C FAM20C FAM72D FAM83D FAM46C FAM53C FAM65C HTATIP2 PARD3B HDAC10 HDAC11 SNRPA1 NT5DC3 SETD1B TDRD12 SCN11A POU5F1 DUSP16 CYP2D6 SPTBN5 DNAJC6 KCNK15 BZRAP1 PCYT1B CELSR1 POU2F3 COX6B2 B3GNT6 IL12RB1 TBXA2R SETD1A CELSR2 PHLDB1 ANKFY1 AMIGO1 B3GNT7 MED12L AHCYL1 ZSWIM6 UBE2D1 MED13L TPRG1L RSC1A1 GALNT6 PTDSS2 SREBF2 ABCC10 PLXND1 TRERF1 RASSF4 HPCAL1 PCNXL3 MTSS1L GALNT2 PLSCR3 AKIRIN2 IL17REL GCN1L1 LDOC1L CELSR3 SESTD1 TUBA1A SELPLG SRD5A2 PCNXL2 OSBPL3 UGT3A1 UGT3A2 TUBB2A DDX12P GALNT7 SLURP1 AKR1C3 KDELR3 PLSCR1 KDELC2 RHEBL1 TUBA1B AKR1C2 POU4F3 MID1IP1 CAPN13 MAD2L1 COX7A2 DNAJC9 NT5DC2 ABHD12 SPTBN2 CYB5R2 OSBPL5 OSBPL7 EMILIN1 SRD5A3 SERHL2 B3GNT3 MRPL39 NUDT12 MB21D1 TMSB10 PHLDB2 HOXA10 CLEC3A ABCC11 RASSF6 AMIGO2 PLXDC2 B3GNT2 TRAFD1 DHFRL1 OSBPL2 UBE2D3 GALNT3 KBTBD4 MRPL52 RAD54B POLR3F MRPL19 SEC11C NUDT21 MRPL35 KBTBD2 CAPZA1 CAPZA2 TRMT11 HRSP12 PTDSS1 MRPL13 PHLDB3 EMILIN2 B3GNT9 DNAJC4 TNRC18 MRPL47 MB21D2 RASSF7 KBTBD6 MRPL21 MRPL32 GALNT1 CYP2U1 MRPL48 ABHD10 KBTBD7 TRMT12 NUDT22 KDELR2 PEX11G MRPL40 ZSWIM3 LMAN2L MRPL51 MRPL54 DUSP19 MRPL46 ZC3HC1 CALML5 MRPL53 ERLEC1 MRPL20 ARPC5L BOLA2B B3GNT1 CIAPIN1 MRPL18 TRMT1L CD2BP2 MRPL50 MRPL22 MRPL27 COX6B1 MRPL33 MRPL11 VSTM2L SEC24D OSBPL9 DYRK1B PCYOX1 DYRK1A TOPBP1 FAM83G FANCD2 NDUFA1 NDUFA8 NDUFA7 NDUFA9 NDUFA4 NDUFA5 NDUFA2 NDUFA6 RAVER1 TROVE2 HOXC10 CAMTA2 PMEPA1 TAMM41 PARPBP HOXC11 DTNBP1 PRR23C PODXL2 POLR2A UNC13A PCDH19 ENTPD1 DNAAF1 PCDH17 SCARF1 ANKS4B DLGAP4 POM121 DNAH10 ADHFE1 SHKBP1 NEURL4 SUPT5H ENTPD8 NEDD4L MFSD12 CC2D1A CDC14A PGM2L1 PPP3CB SEPHS2 TGFBR3 POLR1B LONRF2 CARD14 MAPK12 PPP3CA MFSD10 MAP3K8 GTPBP3 BMS1P5 MAP2K3 PTPRN2 ENTPD2 MAP2K7 MAP3K6 PDCD11 POLR3A KLHDC4 POLR1A POLR1E ANKS1B CRYBB3 MAPK15 TGFBR2 SUPT3H ANTXR2 SFMBT2 CDCA7L MAP4K4 CDC25B DLGAP5 ERCC6L PCDH20 SEC61G PRKAA2 GRPEL2 RAD51C CDC25A CHAF1B CHAF1A TUBA1C SPTSSA SPTSSB CDC14B C2CD4A MAPK13 ENTPD3 PCDH10 RB1CC1 SCPEP1 KLHDC2 PDCD10 N6AMT1 RSPRY1 PRMT10 ENTPD6 MAP2K4 MAP4K3 UBE2Q2 PPPDE1 MAPK10 RABEP2 PSMB10 MAP3K5 GRPEL1 MAP2K1 PPP1CA RHBDL3 LRRC8B SBDSP1 DHRS13 CRNKL1 N6AMT2 SDHAF1 DOLPP1 PRKAA1 GTPBP2 PTPDC1 TMED10 APOLD1 TGFBR1 POLR2K ADAM22 ADARB1 ADAM10 AMOTL2 AMOTL1 MYO15A MYO15B MBOAT7 ACVR1B ACTR3C SACM1L MBOAT2 ACVR2A HAVCR2 PARD6G MGAT4B GNPTAB MGAT4A STAMBP NMNAT2 unknown CCDC92 BCORL1 SCARB1 ARPC1B PTGER3 CCDC78 PTCHD2 CCDC40 SDHAP2 CCDC30 SNAPC4 DPRXP4 MAP7D3 UBQLN2 USHBP1 RHBDF2 ST8SIA4 SHANK2 CLASRP CCDC64 HECTD3 MBTPS1 SDCBP2 LRSAM1 CCDC57 MAN2B2 MAP7D1 CCDC97 MAN2A2 TXNRD3 HECTD2 CCDC47 ADRA1A MAN2A1 PDZRN3 SNAPC2 CCDC86 ST8SIA1 SEMA3F CD3EAP UBQLN4 POLR3H AHNAK2 FCHSD2 ADRBK2 PTCHD1 ASPHD1 SHANK3 KCNAB2 PMFBP1 VSTM2A TMSB4Y RECQL4 ST8SIA6 PKMYT1 CDC25C TMSB4X SHCBP1 NUSAP1 CCDC99 SPRED2 DHCR24 CADPS2 DHTKD1 NDUFS4 NDUFB1 RANBP6 CCDC18 CCDC34 RANBP1 NDUFS7 NPLOC4 CCDC72 NCKAP5 CRABP2 HRASLS EFEMP1 SCUBE3 RNASEL CCDC19 LRRC8C SUCLG2 DOCK10 RBM12B CTNNA1 MRPS10 EIF2AK3 MAN1A2 RANBP2 MRPS14 ARL6IP6 TXNDC5 ARL6IP5 NDUFS6 NDUFB4 GABPB1 CCDC14 ANKRA2 FERMT2 MRPS22 PPIAL4F TXNDC9 CCDC43 PPP1CC LRRCC1 PMS2CL HECTD1 SAMD13 TXNRD1 SCARB2 DOCK11 EIF2AK4 SNAPC5 SUCLG1 ADRBK1 NHLRC3 OBFC2A MRPS31 NECAP1 PSMD10 MRPS11 MRPS15 NDUFS8 NDUFS3 NDUFV2 MRPS24 POLR2C NDUFS1 NDUFB2 MRPS17 MRPS36 PSMD14 NDUFS5 NDUFB8 SNRPB2 NDUFB5 ARL6IP1 NDUFB3 NDUFB6 BDKRB2 CCDC48 SCUBE2 SAMD12 CCDC93 MAN1A1 SAMD11 OGFRL1 VWA5B1 VWA5B2 TOP1MT POTEKP OTUD7B STRADA PRUNE2 PRKAG2 SPATA2L CACNB4 TNFAIP2 LRTOMT CPAMD8 GABBR1 EXOSC3 SORBS1 PCDHB6 PCDHB9 MIR3648 SRGAP1 CSNK1D PCDHB3 ZDHHC8 SORBS3 PRRC2B SPOCK2 PRRC2A ZDHHC3 PRIC285 TMEM80 L2HGDH SYNPO2 GTF2IP1 PIP4K2A PCDHB5 CGREF1 EXOSC5 IRF2BPL PI4KAP1 RPUSD1 FP15737 PCDHB2 PI4KAP2 CALHM2 TMEM64 CRYBG3 CKMT1A CKMT1B NOS1AP SPOCK1 SLITRK4 SEMA6A BAHCC1 ADRA2C MTHFD1 VSIG10L FUNDC1 DEPDC1 TMEM40 SRGAP2 SNRPD1 TMEM60 EXOSC9 EXOSC8 TMEM48 NCAPD2 TMEM97 POLR2G CDKN2A MTHFD2 SEMA4A CYBRD1 MUM1L1 MFSD2A SCAND3 TMEM18 GABBR2 SECTM1 GAPVD1 LMBRD1 JHDM1D TMEM59 TMEM66 ZCCHC4 ZDHHC4 RABEPK WDR45L SEMA3E TMEM62 COQ10B SRGAP3 SLITRK6 RPUSD2 DEPDC4 TMEM85 ANKRD5 U2SURP ZCCHC9 ANKRD9 UQCR10 PIK3AP1 TMEM99 DHRS7B EXOSC1 SNRPD2 TMEM56 PDXDC1 UQCR11 NUCKS1 TMEM65 NDUFC1 CDKN2B CDKN1A NPEPPS SCNN1B PCDHB4 DOPEY1 MSTO2P NOTCH3 NOTCH1 GNPDA2 MCART6 MCART1 ACADSB PRKACA RTN4IP1 AGPHD1 PRKACB MAGEF1 AGTRAP CACYBP PSTPIP2 CAMK2B ZFP36L2 C21orf15 C1orf220 C10orf82 C6orf141 C10orf10 IGF2BP3 C17orf99 SAMD4B R3HDM2 C12orf51 CHRNB1 METRNL C22orf25 C21orf62 POLR2M SAMD4A C20orf54 PECAM1 CAMK2A C19orf55 C17orf63 PIP5K1C C6orf132 C16orf55 C17orf70 R3HDM1 C19orf12 C1orf130 C1orf116 SORCS2 CDKN1C C17orf56 DENND3 C10orf26 C1orf212 SCAMP4 SEMA4C C12orf35 GPRC5A RALGDS C17orf65 C21orf56 R3HDM4 C11orf68 C17orf28 C19orf54 C16orf13 MANEAL PPFIBP1 C14orf43 SEMA6C CAMKK2 C1orf115 PPFIBP2 C19orf66 C14orf25 ECHDC2 CREBBP C19orf31 MAPRE3 C9orf167 PIK3C2B C12orf65 C15orf39 C17orf62 C22orf29 C10orf58 C20orf96 C14orf49 DBNDD1 C19orf80 C1orf226 C6orf192 C11orf88 CHRNB4 C10orf81 C12orf24 C11orf82 C17orf53 C12orf34 MAPRE2 GPCPD1 FRMD4B HIST3H3 C16orf59 C19orf48 C1orf135 C15orf42 C18orf54 NCAPG2 C11orf73 SNORA3 C16orf87 C13orf27 C1orf112 C9orf140 C11orf31 C19orf10 C16orf91 C9orf142 C11orf83 C9orf152 C1orf201 C6orf225 C12orf57 BMPR1B C21orf33 C9orf103 C12orf32 CN5H6.4 C15orf34 C12orf61 SPOCD1 C17orf82 CHRNA5 ECHDC1 C15orf23 C15orf48 C16orf80 CDKN2C CUEDC1 GOLGA4 C12orf26 SNORA8 PCMTD2 PCMTD1 ANKMY2 HMG20A C11orf52 C6orf130 GOLGA5 C22orf28 C14orf80 C15orf29 C12orf29 C10orf28 C21orf91 HIST4H4 ELMOD2 SCAMP1 C16orf52 C20orf20 C12orf75 C19orf63 C6orf165 GOLGA2 ZFP36L1 C15orf44 GOLGA1 C1orf106 C21orf63 C16orf58 C18orf32 C19orf79 C15orf24 C12orf23 NANOS1 C17orf58 C11orf71 ELMOD3 C6orf170 HENMT1 ENOPH1 C11orf92 C6orf168 SCAPER C15orf17 C20orf43 C11orf67 C19orf53 C9orf163 PSENEN C11orf10 C1orf123 C11orf84 C21orf59 C11orf95 C17orf61 C17orf81 C12orf62 C6orf125 C11orf63 LRRFIP2 DNMT3A C16orf45 S100A14 S100A10 S100A16 S100A11 MYEOV2 S100A13 DEPTOR TOMM40 PACSIN3 CACNG8 CACNG4 CACNG1 MAGEE1 JAKMIP3 JAKMIP1 SPATS2L HMBOX1 ATP13A1 ATG16L2 ATP13A2 ATG16L1 ATP13A4 ATP13A5 ATP13A3 KCNMB3 EMX2OS HP09025 MCOLN1 MICALL1 BCL2L15 MIR663B MIR663A BCL2L11 PRKCSH SCNN1G KCNMA1 CAMK1D CHAMP1 MCOLN3 KCNMB4 SNORD8 CHCHD2 CHCHD1 MANSC1 PYCARD GABRG3 TMEM9B CHMP1B MESDC2 DDRGK1 BCKDHB ARMCX3 ORMDL1 SAMHD1 SNORD9 PTTG1IP GPRC5C CHMP2B SETMAR ARMCX5 CCHCR1 NSMCE2 RPL26L1 ORMDL2 ORMDL3 NUDCD2 WASH7P TOPORS MAGEH1 MAGED1 CORO1A CORO2A ATXN7L1 D2HGDH MOSPD3 MARCH8 HMGXB3 ATXN7L2 AMDHD2 TAS2R30 PITPNC1 COMTD1 UQCRC2 EHHADH UQCRC1 MOSPD1 OGFOD1 U2AF1L4 SIGLEC8 EPS15L1 C2orf27A SULT2B1 TCP11L2 TPD52L1 EPB41L1 SLC35F3 CD163L1 GLT25D2 ITGB3BP EPB41L2 MEIS3P1 TPD52L2 SLC35F5 SULT1A1 STAG3L1 STAG3L2 DROSHA CORO1C ATP6V0B TAX1BP3 SLC16A4 HOMER2 SLC14A2 EP400NL SLC12A7 SLC15A2 SEC14L5 SLC23A2 SLC13A5 SLC36A3 B3GALT5 SLC22A4 SLC29A3 SLC25A6 SLC13A3 SLC7A14 SLC26A8 SLC24A4 CASKIN1 CASKIN2 SLC27A4 SLC39A4 SLC16A1 SLC29A2 SLC24A6 ZNF512B MCM3AP GRAMD4 SLC40A1 B4GALT1 SLC35E2 ZNF385A SLC35A4 N4BP2L1 SLC2A13 SLC41A2 SLC35E3 SLC25A1 SLC27A5 SLC45A4 SLC4A11 SEC14L1 SLC12A9 SLC27A1 SLC19A1 SLC24A3 SLC20A1 SLC37A2 SLC29A1 SLC26A2 SLC36A2 SLC15A1 ZNF826P SLC17A7 SLC16A5 SLC43A1 SEC14L2 TFCP2L1 SLC36A1 SLC26A4 HMGCS2 SLC7A13 SLC6A11 SLC7A11 CEP57L1 SLC16A6 SLC44A3 GRAMD2 SLC24A1 SLC39A8 SLC16A3 SLC12A8 ZNF204P SLC4A10 SLC16A9 SLC12A1 MARCKS SULT1C4 ZNF804A SLC35B3 SLC38A1 ZNF518A B4GALT4 SLC35A5 SLC30A7 SLC39A1 SLC30A8 N4BP2L2 SLC11A2 RPL36AL SLC20A2 GRAMD3 SLC30A9 SLC35A3 SLC38A2 PDE4DIP SLC39A7 SLC37A1 SLC31A2 SLC10A7 SLC39A6 B3GALTL ZNF780A SLC37A3 ZNF518B DPY19L4 SLC31A1 SLC35B1 SLC30A5 HOMER1 SLC35A1 PGRMC1 SLC12A4 SLC30A4 SLC44A1 RWDD2A RWDD2B CBFA2T3 CBFA2T2 FAM71F2 FAM71F1 KRT8P41 TSPAN10 TSPAN33 ATP6V0C TSPAN14 KATNAL1 TSPAN15 TSPAN31 ATP6V1D CERCAM TSPAN13 TSPAN12 PITPNM2 TNFSF14 L3MBTL1 POLR2J4 ZFYVE28 MICALCL TNFSF15 SERINC5 ZFYVE26 PLA2G10 POLR2J3 SERINC2 RRPL13L CYP4F22 TIMM17A TRNA_Ile TIMM17B TNFSF10 PITPNM1 ZNF280D SERINC1 SERINC3 PLA2G15 ZFYVE19 TBL1XR1 RSL24D1 SLC35D1 C1GALT1 SLC35C1 AGXT2L1 FAM86B1 FAM95B1 FAM155B FAM157A FAM193A FAM215A FAM153B FAM127B FAM102A FAM110B FAM176B FAM123B FAM131A FAM116B FAM161B FAM127A FAM134B FAM175A FAM164A FAM208B FAM117B FAM100A FAM117A FAM213B FAM120A FAM189B SOWAHB FAM179A FAM124A FAM196A FAM131B FAM111B FAM198B FAM105A FAM100B FAM129A FAM167A FAM113B FAM179B FAM135A FAM172A FAM82A2 FAM190A FAM169A FAM200B FAM104B FAM102B FAM214A FAM111A FAM91A1 FAM162A FAM192A FAM211B FAM92A1 FAM154B SOWAHA FAM173B FAM176A FAM18B1 FAM19A5 PABPC1L HSPA12A OSTalpha RNF144A GAL3ST4 CYP46A1 TBC1D17 ZC3H12A ZC3H12B TSC22D4 CYP20A1 TBC1D16 PPP1R21 ZNF37BP COL14A1 ST3GAL5 ST3GAL3 TBC1D25 RNF144B TBC1D24 EHBP1L1 AFG3L1P STXBP5L RPS2P32 RASL10B COL13A1 COL27A1 CYSLTR1 HIST1H3I AKR1B10 ST6GAL1 GAL3ST2 COL18A1 ST3GAL1 RASL11B HIST1H4I CYP39A1 ST3GAL4 ZC3H11A TSC22D1 METTL14 TBC1D23 DNAJB14 METTL15 ADIPOR1 METTL17 RASL11A DNAJB11 DONSON METTL18 COL28A1 COL17A1 TACSTD2 FAM123C FAM122C RTN4RL2 FAM129C FASTKD1 SOWAHC FAM110C MAGOHB GATAD2B GATAD2A FLJ10661 ZC3H12D FLJ36777 ALDH1B1 DNAJC22 PTPLAD2 FDPSL2A CD300LG FLJ42102 FLJ14107 FLJ42627 KBTBD12 FLJ43663 FLJ45513 FLJ32756 FLJ12825 FLJ00310 UNQ6975 FLJ14186 GALNTL2 ALDH3A1 FLJ12334 CLEC18B FLJ40194 FLJ00285 FLJ33360 CLEC18A CLEC11A BHLHE40 GALNTL1 GALNT10 CLEC16A FLJ00322 FLJ23867 UNQ2790 FLJ45340 FLJ14346 KBTBD11 GALNTL4 SDR42E1 FLJ22184 PLEKHF1 FLJ46481 FLJ42709 CREB3L1 HSD17B2 SUV39H2 PKHD1L1 BHLHE41 GALNT14 FLJ46906 OSBPL10 DNAJC17 HSD17B6 KAZALD1 GALNTL6 UGT2B10 HSD17B1 FLJ33065 DNAJC15 FLJ34503 FLJ27352 CYP4Z2P MUDENG ZCWPW1 DNAJC25 COMMD6 SPTY2D1 COMMD9 SAP30BP FLJ23152 PTPLAD1 PLEKHF2 DNAJC19 ALDH5A1 HSD11B2 UGT2B15 UGT2B17 HSD17B4 COMMD4 ALDH1A1 COMMD1 ALDH7A1 EFCAB11 DNAJC28 COMMD2 ALDH1A3 DNAJC10 TRMT112 FLJ10038 SERTAD2 SERTAD1 PGLYRP2 ZFAND2A FAM86EP JA040723 JA040725 NDUFA12 HEATR7A HEATR5A KIAA0195 KIAA1107 KIAA1467 KIAA0922 KIAA2012 KIAA0182 ZSCAN12 KIAA1875 KIAA1530 KIAA1549 KIAA2018 KIAA1751 GPR172B KIAA1614 TNFRSF9 KIAA0913 KIAA1522 GLTSCR1 PLEKHA7 KIAA1324 KIAA1609 MAP3K11 CABLES1 SH3GLB2 KIAA0247 KIAA0430 THSD1P1 TBC1D9B BLOC1S3 KIAA0284 PPP2R5E KIAA0754 KIAA1737 SDR39U1 KIAA0930 SIGMAR1 KIAA2022 KIAA1671 KIAA0664 KIAA0889 KIAA0226 KIAA1958 REXO1L1 KIAA1804 SLCO4A1 KIAA1199 RPS6KA2 KIAA1653 PLEKHA6 SLCO3A1 LDLRAD3 LDLRAD2 PPP1R1B NUP62CL KIAA1524 KIAA0101 AZGP1P1 ANXA2P1 METTL7A SDR16C5 KIAA0040 EXOC3L4 ANXA2P3 RPS6KA5 KIAA1704 PLEKHA8 MAP3K13 SUPT4H1 KIAA0232 KIAA0895 TWISTNB MMADHC PLEKHA2 BLOC1S1 KIAA1257 KIAA1456 SUPT16H CTDSPL2 RPS6KB2 KIAA0391 KIAA1024 TRAM1L1 BLOC1S2 RPS6KA3 KIAA0355 LRRC37A KIAA1147 KIAA1377 RUNX1T1 DYNLRB2 FAM86DP FAM86HP NDUFAF1 NDUFAF3 AL832384 CCDC120 CCDC165 RNASE10 MGC4836 AL833150 AL137655 COLEC12 SUGT1P3 POU2AF1 AL832007 NCKAP5L TRMT61A CCDC124 PLEKHN1 PPP2R2D TXNDC16 AL137535 CCDC106 PLEKHH3 RANBP3L PLEKHD1 AL832439 PLEKHH1 AL360260 CCDC104 C1QTNF6 CCDC150 AL157440 TBC1D3C ARHGDIA CSNK2A2 RNPEPL1 AL832797 PPP2R2C ABHD12B HRASLS2 MGC2889 CTNNAL1 TXNDC17 CCDC167 AL832163 SLCO4C1 TMSB15A CCDC160 ARHGDIB ABCC6P2 LAPTM4A CCDC132 CCDC126 AL355711 CCDC127 TXNDC12 SERPINI1 PPP1R1C CCDC111 AL833455 AL110181 PPP2R5C CCDC134 CCDC121 OSBPL1A CCDC149 EFCAB4B PPP1R3D CSNK2A1 CCDC71L ANAPC10 TIMMDC1 ANAPC11 POU5F1B ANAPC13 PLEKHH2 PPP2R3C ZBTB8OS AL832909 C1QTNF1 KRTCAP2 DYNC2H1 SNAR − A1 NDUFAB1 TRNA_Tyr ZCCHC14 CYP2G1P TMEM236 TMEM136 TMEM101 RNASET2 ZCCHC24 DNM1P35 PCDHB18 TMEM212 TSNARE1 PCDHB13 PCDHB12 TMEM213 ZDHHC11 TMEM201 ZDHHC15 PCDHB10 TMEM131 SNRNP70 GUSBP11 PCDHB16 ANKRD36 TMEM175 TMEM231 HSPBAP1 TRAPPC2 NR2C2AP ZDHHC14 ANKRD11 KLHDC8B HERC2P7 TMEM180 ANKRD52 TMEM198 TMEM102 NSUN5P2 ZDHHC23 ZKSCAN1 PCDHB11 ANKRD16 TMEM105 NSUN5P1 MTHFD1L NEURL1B SCGB2A2 TMEM139 TMEM107 TMEM147 ANKRD37 SNRNP27 HIST1H4J TMEM182 TMEM154 ZCCHC11 HIST1H3J TMEM168 PLEKHG3 TMEM203 ZDHHC13 ANKRD49 TMEM135 TMEM123 ANKRD32 TMEM242 POLR3GL ANKRD46 TMEM223 MORF4L1 ZDHHC20 PRPSAP2 TMEM128 TMEM205 ZCCHC10 TMEM111 ZKSCAN4 TMEM165 TMEM208 SNRNP25 TMEM141 TMEM178 ANKRD12 HERC2P9 HERC2P3 HERC2P2 HERC2P4 ANKRD50 AF339817 AF090939 AF079515 AF289551 DL492557 BAIAP2L1 DL492607 AF283776 AF086132 PHACTR4 FLYWCH1 HNRNPA0 FOXRED2 AK097527 AX747261 AX747158 AK095700 AK128400 AX748268 C20orf112 AX748239 AK127460 AK055254 AX748265 BX647715 BX648961 BX641154 AK056486 AX748195 AX748283 AK093203 AX746599 AK094692 AX747521 AX746635 AK126380 AK302511 AX747080 C10orf113 AK125684 AX721280 AK094352 AK056556 FBXW4P1 AX747592 AX746575 AK125769 AX746834 AX748371 AK127190 C14orf105 AK096230 AK131520 AX747721 AX746787 BX641143 AK125437 AK056817 BX537710 AK126253 BX648496 AK124523 AX746871 AK055666 AK123617 AX747333 AX747182 AK022382 AX747756 AK125858 AX747648 AX747586 AK096475 AX746755 AX746699 AK128346 AK123177 AX748196 AK057887 AX747507 AK075019 AK098263 AK123543 AX748181 AX746504 AK027091 AK056253 AK124578 AK124439 AK091729 AK296966 AK023304 AK023515 AX746877 AK096249 AK125412 PRKAR1B BX161428 AX746617 AX746771 AK095045 AX747766 AK123816 AK127026 AK128777 AK055876 AX746851 AX746839 AK095583 AK097289 AX748420 AK130294 AK024248 BX537921 AK128559 AX747140 AK131347 AX746964 AX747444 AK096219 AX747238 AX746765 C14orf132 AK098727 AX746734 AX747730 AX746995 BX648270 AK022914 AX747325 PLEKHM1 RAPGEF1 C17orf103 AX747742 PLEKHM2 AK130329 C10orf137 AK128619 UBE2CBP AX748175 AK309255 AX747590 AK023288 AK055017 AK125301 BX537481 AK095151 AK000470 CCDC85B AK097283 AK056524 AK090700 BX648826 AK055918 AK124832 AK125489 AK097370 AK130076 AX748417 AX748308 AX747598 AK094577 AK091996 AK127555 AX746823 AK128153 AX748330 AK311282 AK096982 AX748385 AX746638 AK304826 AK123891 AK310228 AX747628 AX748411 ZMYND17 AK097803 AK023178 AX748285 AK097143 CTDNEP1 AX747795 AK000798 POM121C AX746867 PLEKHM3 AK125257 AX748125 AX748135 AK094963 AX747191 AK074469 AX747537 GUCY1A2 AK127846 AX747826 SH3BGRL AK127378 AB074160 AX747913 BX649128 RAPGEF4 CNTNAP2 CCDC88A AK125288 AK026965 AX747684 AX747440 SCGB1D2 AK125429 AK093600 AX747630 RAPGEF5 PCOLCE2 EIF4EBP1 SNORA21 AX747493 SNORA75 SNORA84 SNORA18 SNORA48 SNORA34 SNORA70 SNORA24 SNORA10 SNORA52 SNORA67 SNORA54 SNORA22 SNORA27 SNORA63 SNORA38 SNORA40 SNORA79 SNORA11 SNORA28 SNORA68 SNORA37 SNORA49 SNORA31 C10orf112 SNORA20 C14orf159 SNORA72 SNORA80 AK126266 C21orf119 SNORA65 CCDC74A AX747104 GUCY1A3 AK095117 AK056897 HIST1H4L AK056683 AX748345 AK125397 AK309286 AK094156 AK126744 SNORA50 AX747981 GUCY1B3 RAPGEF3 AK056252 BX538254 AK056098 AK126307 ARHGDIG AK127609 AK027179 AX747192 PRKAR2B C14orf101 C20orf177 CSRP2BP C14orf119 ARFGAP2 PRKAR1A AK098012 SNORA12 AK021570 AK130759 AK311660 AK289390 C14orf129 SNORA30 AX747544 CCDC28A AK027541 SMPDL3A AK092541 AK093871 AB074166 TRAF3IP2 MRPS18A C14orf142 SNORA42 AK307134 AK096279 SNORA46 GADD45A ADAMTS4 TOMM40L ADAMTS1 LAMTOR3 DYNC2LI1 DYNC1LI2 SND1 − IT1 MAGEA10 PYROXD2 CCDC88C ARHGEF3 ARHGEF4 TMEM63B TUBGCP6 CRAMP1L ARHGEF7 PDXDC2P CHCHD10 RHOBTB2 CENPBD1 ARHGEF1 KHDRBS3 RASGRF1 RHOBTB1 HIST1H3F HIST1H4F DEPDC1B SNORD10 TMEM38B SCARNA4 SNORD97 SNORD94 SCARNA1 SNORD89 SCARNA6 TMEM45B CARHSP1 DSCAML1 RARRES3 ARHGEF6 TMPRSS4 MTERFD3 TMEM50A TMEM50B TMEM39A TMEM30A MRPS18C DCUN1D1 TMPRSS2 ARHGEF2 TUBGCP5 SCARNA9 SCARNA7 TMPRSS3 SCARNA5 TMEM87A TMEM45A TMEM14B DCUN1D3 SCARNA3 TMEM86A MTERFD2 BC042385 BC040333 BC035181 BC090058 BC047364 BC040563 BC019672 BC041007 BC160930 BC148242 BC013821 BC051736 BC045578 BC045192 BC053669 BC026354 BC045779 BC029255 BC043227 BC032415 BC024027 BC035084 BC035187 BC034929 BC043213 BC037540 BC036850 BC016143 BC043205 BC034268 BC035179 BC047626 BC047625 BC071667 BC022892 BC035176 BC031979 BC048427 BC045769 BC036311 BC045805 BD495725 BC040646 BC038742 BC041025 BC006435 BC045784 BC062350 BC042590 BC108660 BC045560 SYNGAP1 BC045731 BC021061 BC047513 BC018860 BC040358 BC040577 BC041030 BC041449 BC038464 MYBBP1A BC037357 BC038455 BC033260 MAPKBP1 BC014180 BC035340 BC020935 BC043223 BC014606 BC070093 BC033373 BC011243 BC131768 BC016972 BC015784 BC132896 BC052775 BC032795 BC101234 BC039389 BC034827 BC039356 BC034795 BC046476 BC027846 SEPSECS BC038570 BC073897 BC070371 BC067244 BC150535 BC095475 RAP1GAP BC047609 BC010186 BC015429 BC042448 PCDHAC2 CACNA1H RACGAP1 HLA − DMA OIP5 − AS1 CACNA1D RASGRP3 DENND2A GOLGA8B SIGLEC10 EGFEM1P DENND4B GHRLOS2 DENND1A GOLGA8A CR936688 CAMSAP3 DENND5B HIST1H4K CR627135 GOLGA7B HIST1H4B HIST1H3E HIST1H3B HIST1H1B HIST2H4A HIST1H3A HIST1H4A SNORA2A SNORA5A ADORA2B ONECUT2 IRAK1BP1 CR936711 ADORA2A HIST1H4E MIS18BP1 ARHGAP5 DENND1B RG9MTD2 UQCRFS1 TMEM14C RUNDC3B CEACAM1 CEACAM7 CEACAM5 ATP6V0A4 ATP6V0A1 ATP6V0E2 ATP6V1E2 THUMPD1 ATP6V1E1 HNRNPH2 HNRNPH1 THUMPD3 PRICKLE1 DQ573539 LOC79015 DQ592442 DQ595103 SLC22A20 SLC25A35 LOC23117 SLC25A45 DQ574721 PRICKLE3 SLC25A34 SLC25A30 SMARCA4 SLC25A29 SLC26A11 DQ576756 GTF2IRD1 SLC25A23 SLC25A25 DQ598910 DQ587763 SLC25A22 DQ596646 SLC25A10 SLC25A43 KREMEN1 LOC90834 DQ571357 SLC38A10 SLC39A11 SLC16A13 DQ570835 KREMEN2 NSMCE4A SLC25A48 DQ593686 DQ583756 DQ577420 HIST1H3H HIST2H3C DQ786304 HIST1H1D HIST1H3C HIST1H3D HIST2H3D HIST1H4D HIST1H4H SLC25A20 SLC25A12 SLC16A10 SLC16A14 DQ574505 SLC25A33 SLC22A17 HIST1H4C SMARCE1 DQ593252 SLC25A32 SLC25A38 HIST1H1C GTF2IRD2 DQ586540 LOC81691 LOC90246 RUNDC2C EDARADD KRT18P55 CROCCP2 CROCCP3 TP73 − AS1 ATP6V1C2 PCDHGC5 FW339974 SMARCC2 FW340097 HM358993 FW339973 HIST1H3G RABGGTB EPM2AIP1 ZMYM6NB FAM160B2 FAM160A2 FAM189A2 FAM103A1 FAM114A1 FAM177A1 ATP6V1G1 SERPINE1 PPP1R13L SERPINB6 SLC35E2B RNF138P1 GRAMD1A SLC7A5P2 SERPINB5 SERPINA3 TRNA_Leu MACROD1 WDR83OS CTNNBIP1 FAM108C1 ZC3HAV1L CYP4F35P HSD17B13 MIR143HG HSD11B1L HSD17B14 ProSAPiP1 CYP4F30P HS3ST3B1 HSD17B12 GRAMD1C NUDT16L1 COMMD10 KIDINS220 MIR600HG SLC9A3R2 FRA10AC1 SHROOM2 PABPC1P2 LEPROTL1 LINC00485 LINC00174 LINC00092 LINC00311 LINC00202 METTL21A LINC00085 TBC1D22A LINC00472 PPP1R15A PPP1R13B LINC00294 LINC00482 LINC00152 HIST1H2BI TNFRSF21 RPS19BP1 LINC00467 CYP2B7P1 LINC00312 PPP1R14B AKR7A2P1 TNFRSF18 LRRC37A4 SLC4A1AP HSP90AB1 LINC00493 TNFRSF19 LRRC37A2 LRRC37A3 LINC00284 MRFAP1L1 STARD3NL CYP2D7P1 DNASE1L2 B4GALNT4 STAMBPL1 RAD51AP1 B4GALNT2 B3GALNT1 B3GALNT2 NAALADL2 B4GALNT3 TCTEX1D2 RNU4ATAC RNU6ATAC LGALS3BP A1BG − AS1 TNFAIP8L1 TNFAIP8L3 CATSPER2 CDC42BPA RALGAPA2 UBE2Q2P1 LOH12CR1 CDC42EP4 CDC42EP2 UBE2Q2P2 TMED10P1 CDC42EP5 TRAPPC11 SELENBP1 SLC7A6OS CDC42SE1 HLA − DRB5 CCDC109B PPAPDC1B CCDC102A SH3BGRL3 MAP1LC3B RAPGEFL1 SH3PXD2B GOLGA6L9 SH3BGRL2 MAD2L1BP ADAMTS17 ADAMTSL2 ADAMTS19 ADAMTS13 ADAMTS15 ASAP1 − IT1 HLA − DQB1 MTRNR2L5 MTRNR2L6 MTRNR2L3 TMPRSS12 ANKRD13B CDC42BPB TGFBRAP1 MTRNR2L4 TMEM184B RAB11FIP3 ANKRD18A RAB11FIP2 TMEM229B ARHGEF37 CDK5RAP2 ZDHHC8P1 PCDHB19P ARHGEF10 TMEM120B TMEM120A MTRNR2L7 TMEM185A ANKRD19P HIST1H2AJ HIST1H2BJ SRGAP2P2 SCARNA23 TMEM194A SCARNA18 XRCC6BP1 TMEM191A ANKRD34B SCARNA20 TMEM106A TMEM194B SCARNA12 SCARNA22 ANKRD34A SCARNA14 ARHGEF25 ANKRD30B TMPRSS13 ANKRD30A TMEM106B TMEM167B TRAPPC6B TMEM161B SCARNA10 ARHGEF38 TMEM167A TMEM126A ARHGEF40 MTRNR2L8 ANKRD13A MTRNR2L1 TRNAU1AP CACNA2D2 CACNA2D3 PPARGC1B PPARGC1A RASGEF1B ARHGAP32 ARHGAP28 SNORA74B ARHGAP33 SNORA74A APOBEC3F ARHGAP39 PLEKHM1P ARHGAP26 ARHGAP17 ARHGAP35 HIST1H2AL HIST1H2BL SNORA71B SNORA14B SNORA38B SNORA11B SNORA59B ARHGAP18 TMEM150C TMEM184C TMEM106C CNTNAP3B SNORA14A ARHGAP10 ARHGAP24 KCNQ1OT1 CEACAM20 − AS1 CRYM LOC619207 LOC283050 LOC283663 LOC400958 LOC150776 LOC440905 LOC731275 LOC149134 LOC283767 LOC728323 LOC285972 LOC440300 LOC728190 LOC644215 LOC285103 LOC284260 LOC283089 LOC646762 LOC283693 LOC399753 LOC644961 LOC643988 LOC613037 LOC386597 LOC731424 LOC145663 LOC151475 LOC387723 LOC728558 LOC652276 LOC648987 LOC145820 LOC221442 LOC646324 LOC399744 LOC649305 LOC148709 LOC144481 MARCKSL1 LOC387647 LOC493754 LOC643837 LOC283070 LOC641298 LOC646329 LOC595101 HNRNPUL2 LOC645431 LOC730091 LOC730101 LOC338817 LOC284865 LOC389791 LOC440354 LOC283922 LOC389634 LOC284454 LOC285484 LOC400043 LOC643770 LOC113386 LOC642846 LOC283335 LOC388796 LOC284379 LOC728758 LOC283174 LOC344595 LOC401074 LOC286467 LOC282997 LOC729737 LOC286002 LOC729970 HIST2H2BF LOC541471 APOBEC3B RNASEH2A SNORA71D SNORD15B LOC283104 LOC401397 LOC401109 LOC440925 LOC375295 LOC440461 LOC401093 LOC285419 HIST1H2BF LOC254559 LOC338758 LOC400657 SNORD15A LOC729852 LOC145837 LOC440335 LOC550112 LOC645212 LOC374443 LOC645249 LOC729013 LOC644656 LOC494150 LOC400027 LOC645513 APOBEC3D HIST2H2BA HIST1H2AK HIST1H2BB HIST1H2AB HIST1H2BE HIST3H2BB HIST2H2AB HIST1H2AE HIST1H2BK RNASEH2C HIST2H2AC CALCOCO1 PCDHGB8P DPY19L2P2 HIST1H2BH HIST1H2BN HIST1H2AH HIST1H2AD RPL23AP32 HIST1H2BD HIST1H2BC TTC28 − AS1 HIST1H2AG HIST1H2BG HIST1H2BO SMARCAD1 − IT2 BRWD1 HIST1H2BM MPHOSPH9 HIST1H2AM MPHOSPH6 MPHOSPH8 C1GALT1C1 ZNFX1 − AS1 UCKL1 − AS1 RAD21 − AS1 TNFRSF12A TNFRSF10A TNFRSF10B TCRBV14S1 LINC00256A HSP90AB4P MCF2L − AS1 CLRN1 − AS1 HNF1A − AS1 TNFRSF10C − AS1 UBAC2 ASMTL − AS1 ANKRD20A2 ANKRD20A4 CTTNBP2NL MTRNR2L10 ANP32A − IT1 ISY1 − RAB43 ARHGAP11B ARHGAP11A HIST2H2AA3 CEACAM22P OK/SW − cl.16 GTF2IRD2P1 FBXL19 − AS1 DNMBP − AS1 NME1 − NME2 MPHOSPH10 LGALS8 − AS1 cytochrome_b POM121L10P CDKN2AIPNL − AS1 FAM13A NCRNA00269 BIVM − ERCC5 GPR75 − ASB3 Metazoa_SRP − AS1 ADARB2 ST6GALNAC4 ST6GALNAC5 ST6GALNAC6 SHANK2 − AS3 IQCJ − SCHIP1 − 1 ERVMER34 ANKRD20A5P ANKRD20A9P hCG_1997999 hCG_1980662 C17orf76 − AS1 CSGALNACT2 HSPE1 − MOB4 NEDD8 − MDP1 − P2RY11 PPAN DNAJC27 − AS1 − AKAP2 PALM2 ANKRD20A11P LOC100133161 LOC100506385 LOC100131257 LOC100133091 LOC100293516 LOC100505933 LOC100294362 LOC100507117 LOC100129617 LOC100287225 LOC100507424 LOC100506688 LOC100190938 LOC100190939 LOC100132247 LOC100616668 LOC100652768 LOC100131551 LOC100134713 LOC100287314 LOC100133331 LOC100128682 LOC100190940 LOC100506241 LOC100130331 LOC100131096 LOC100128531 LOC100130581 LOC100507178 LOC100506462 LOC100507173 LOC100287722 LOC100289019 LOC100129269 LOC100287792 LOC100093631 LOC100129550 LOC100527964 LOC100131193 LOC100506990 LOC100271836 LOC100216546 LOC100335030 LOC100190986 LOC100506599 LOC100505676 LOC100506123 LOC100272216 LOC100132077 LOC100506660 LOC100506334 LOC100506730 LOC100505576 LOC100130899 LOC100268168 LOC100128881 LOC100129034 LOC100505989 LOC100128191 LOC100049716 LOC100133669 LOC100288637 LOC100499405 LOC100289361 LOC100131496 LOC100128098 LOC100134229 LOC100130855 LOC100506810 LOC100506783 LOC100129794 LOC100272217 LOC100506013 LOC100289211 LOC100289255 LOC100506779 LOC100130992 LOC100270746 LOC100507495 LOC100506207 LOC100288181 LOC100270804 LOC100505648 LOC100507557 LOC100630918 LOC100294145 LOC100288748 LOC100288428 LOC100505783 DKFZp451F083 TRIM6 − TRIM34 DKFZp761E198 Em:AC005003.4 SNURF − SNRPN DKFZp434J0226 TMED7 − TICAM2 NT5C1B − RDH14 DKFZp434P0216 DKFZp667P0924 C15orf38 − AP3S2 PRR5 − ARHGAP8 DKFZp686L08115 − GTF2A1L STON1 TNFSF12 − TNFSF13 ARMCX5 − GPRASP2 intersectin_1_long_form

DTX2P1 − UPK3BP1 PMS2P11

DTEP L2 P6 L2 DTEP P7 L1 DTEP

Untreated1 Untreated2 DTEP L1 P12 L1 DTEP

Figure B.4: DE-Seq Finds 6943 Signiﬁcantly Diﬀerentially Expressed Genes. Heatmap displaying consensus clustering [153] of all 6,943 genes with a DE-Seq [3] adjusted p-value below 0.1 for any of the 3 possible comparisons (untreated vs. DTP, untreated vs. DTEP, or DTP vs. DTEP). Samples clustered into 3 main clusters by sample type. DTEP L1P7: drug tolerant expanded persisters line #1 passage #7; DTEP L1P12: drug tolerant expanded persisters line #1 passage #12; DTEP L2P6: drug tolerant expanded persisters line #2 passage #6; DTP: drug tolerant persisters.

102 DTEPhccdepmg08 L2 P6

DTEPhccdepmg07 L1 P12

DTEPhccdepmgA6 L1 P7

DTPhccdtpmg05 2

DTPhccdtpmg02 1

Untreatedhccparmg03 2

Untreatedhccparmg01 1 DTP 2 DTP 1 hccdtpmg05 hccdtpmg02 hccparmg03 hccparmg01 hccdepmg08 hccdepmg07 hccdepmgA6 Untreated 2 Untreated 1 Untreated DTEP P6 L2 DTEP P7 L1 DTEP P12 L1

Figure B.5: DE-Seq reveals that untreated cells, DTPs and DTEPs have sig- niﬁcantly diﬀerent expression patterns, with DTEPs being the most distinct. Heatmap displaying Euclidian distance between samples, calculated using the variance stabilizing transformation of the DE-Seq [3] count data, red is 0 distance. DTEP L1P7: drug tolerant expanded persisters line #1 passage #7; DTEP L1P12: drug tolerant expanded persisters line #1 passage #12; DTEP L2P6: drug tolerant expanded persisters line #2 passage #6; DTP: drug tolerant persisters.

103 GNB1

G_beta/gamma/PI3K_(complex)

GMCSF/GMR_alpha/CSF2RB/JAK2_(dodecamer)/SHC/GRB2/GAB2/PI3K_(complex)IL3/IL3RA/CSF2RB_(dodecamer)/JAK2/JAK2/SHC/GRB2/GAB2/PI3K_(complex)

GAB2/PI3K/SHP2_(complex)

PIK3CA IL2/IL2R_alpha/beta/gamma/JAK1/LCK/JAK3/IRS1-2/PI3K_(complex) GRB2/GAB2_(complex) IL2/IL2R_beta/gamma/JAK1/Fyn/JAK3/PI3K_(complex) IL2/IL2R_beta/gamma/JAK1/LCK/JAK3/PI3K_(complex) GAB2

IL2/IL2R_alpha/beta/gamma/JAK1/LCK/JAK3/SHC/GRB2/SOS1/GAB2/SHP2/PI3K_(complex)

IL2/IL2R_alpha/beta/gamma/JAK1/LCK/JAK3/SHC/GRB2/SOS1/GAB2/ERK1-2_(complex) GRB2/SOS1/GAB_family/SHP2/PI3K_(complex)SHC/Grb2/SOS1/GAB1/PI3K_(complex)

IL2/IL2R_alpha/beta/gamma/JAK1/LCK/JAK3/SHC/GRB2/SOS1/GAB2/SHP2_(complex) IL2/IL2R_alpha/beta/gamma/JAK1/LCK/JAK3/SHC/GRB2/SOS1/GAB2_(complex) RAS_family/GTP_(complex) FRS2_family/SHP2/CRK_family/C3G/GAB2_(complex)

GAB2/GRB2/SOS1_(complex) JAK3 IL4/IL4R/JAK1/IL2R_gamma/JAK3/FES/IRS2_(complex)

IL4/IL4R/JAK1/IL2R_gamma/JAK3/IRS1_(complex)

IL-2/IL-2R-alpha/IL-2R-beta/IL-2R-gamma/JAK1/JAK3_(complex)PI3K_(complex)

Insulin_Receptor/Insuli/IRS1/GRB2/SHC/Sos_(complex)IRS2 SOS1 IL2/IL2R_alpha/beta/gamma/JAK1/LCK/JAK3/STAM1-2_(complex) SOCS1 IGF-1R_heterotetramer/IGF1/IRS1/GRB2/Sos_(complex) RAPGEF1

MYC/Max_(complex) IL-7/IL-7Ra/IL-2R-gamma/JAK1/JAK3_(complex) GRB2/SOS1_(complex) IRS1 IL2/IL2R_alpha/beta/gamma/JAK1/LCK/JAK3_(complex) IRF1 PP2A_Heterotrimer_(complex) AKT1 IGF-1R_heterotetramer/IGF1/IRS1/GRB2/Sos/Shc/RACK1_(complex) EPO/EPOR_(dimer)/JAK2_(complex) CRKL PRKCE STAT3_(dimer_)_(complex) PRKCD

CtBP/CBP/TCF1/TLE1/AES_(complex) p53_(tetramer)_(complex) beta_catenin/TCF1/CtBP/CBP/TLE1/AES/SMAD4_(complex)MYC NF_kappa_B1/RelA_(complex) BAD/BCL-2_(complex) ELAVL1 MYC/Max/MIZ-1_(complex) SOCS3 SDC4 CBP/p300/CREB_(complex) STAT1_(dimer)_(complex) NF_kappa_B1_p50/RelA_(complex) MYB RAF1 Nmi/p300/CBP_(complex) MAP2K1 SMAD2/SMAD2/SMAD4/p300/CBP_(complex) BCL2 IL2R_beta/JAK1/LCK/SOCS3/SOCS2_(complex)IKK_complex_(complex) MITF/p300_(complex) STAT5/STAT5/Nmi/p300/CBP_(complex) CBP/p300/ACTR/RXR/CARM1/RAR/retinoids_(complex) p300/CBP_(complex) p300/CBP/RELA/p50_(complex) RXR/PPAR-gamma/PGC1-alpha/SRC1/CBP/p300_(complex) CYCSSMARCC2 MAPK3 SOCS2 NEDD4LEstrogen/ER_alpha/Estrogen/ER_alpha/GRIP1/CARM1/p300(Me)/CBP(Me)_(complex) p300/CBP/Pitx2_(complex) CREBBP RELA NICD/YY1_(complex) PRKCA TCF4/beta_catenin/JUN/DVL3_(complex)JUN/JUND_(complex) NCOR2 MAPK14 SMAD3/SMAD4/p300/CBP_(complex)

TAp73a_(tetramer)_(complex)PPARG/RXRA/fatty_acid/Mediator/Coactivator_Complex_(complex) EP300 RELA/p50/ATF-2/IRF/c-JUN/HMG1/PCAF/CBP/hSWI/SNF_(complex)IFN-beta_nucleosome_(complex) p-KIT_(S741746)/PKC_alpha_(complex) GATA3 ATF2 NEURL JAG2 RXRs/NUR77/BCL2_(complex)RELA/p50/ATF-2/IRF/c-JUN/HMG1/PCAF/CBP_(complex) HES1/PARP1/Cbp_(complex) MAPK1 JUN/Cbp_(complex) RELA/p50/ATF-2/IRF/c-JUN/HMG1/PCAF_(complex)RELA/p50_(complex) JAG1 STAT4_(dimer)/JUN/Cbp_(complex) NECD/Jagged_1_(complex) RELA/p50/ATF-2/IRF/c-JUN/HMG1_(complex) NECD/Jagged_2_(complex) RBPJ/NICD/MAML1-2/p300_(complex) NOTCH1/Jagged_1_(complex) JUN/JUN-FOS/p300_(complex) HIF-1-alpha/ARNT/CREB/p300/JAB1/c-JUN_(complex) NOTCH1_heterodimer/Jagged_1_(complex)NOTCH1_heterodimer/Jagged_2_(complex) EGR1 NOTCH1 NFATc(+)/CBP_(complex) MAPK8 KLK3 Gamma_Secretase_(complex) JUN HES1 NR4A1 JUN/ATF2_(complex) gamma-secretase_complex_(complex) JUN_(dimer)_(complex) IL13 ADAM_10_metalloprotease_(Zn_cofactor)_(complex) IL4 c-JUN/ATF-2_(complex) SNAI1 NOTCH3 dNp63a_(tetramer)_(complex) NTM1/NEC1_heterodimer_(complex) ATF2/c-Jun_(complex)

NOTCH1_heterodimer_(complex) AP1_(complex)

NTM3/NEC3_heterodimer_(complex) SMAD3/SMAD4/JUN/FOS_(complex)JUN/FOS/GR/TRIP6/cortisol/GR_alpha_(dimer)_(complex)NFAT1/FOXP3_(complex)JUN/FOS/GATA2_(complex)AP-1(cJun/cFos)_phosphorylated_(complex) JUN/FOS/NFAT1-c-4_(complex) JUN/FOS/CRTC1_(complex) FURIN JUN/FOS/NFAT1_(complex)JUN/NFAT1-c-4/p21SNFT_(complex)JUN/FOS/TCF4/beta_catenin_(complex) JUN/FOS/RACO-1_(complex)NFAT1-c-4/ICER1_(complex) NFAT1/JUN/FOS_(complex) ELK1 JUN/FOS/ER_alpha_(complex)JUN/FOS/TRIP6_(complex) NFAT1-c-4/PPARG_(complex)cortisol/GR_alpha_(monomer)/JUN/FOS_(complex)JUN/FOS_(complex)

CSNK2A1 NFATC1 FOS

CaM/Ca2+/Calcineurin_A_alpha-beta_B1_(complex)

Calcineurin_(complex) SRF ELK1/SRF_(complex) SRF(dimer)/MAL(dimer)_(complex) CREB1

GSK3B

PRKACA

cNAP-1_depleted_centrosome_(complex) centrosome_containing_phosphorylated_Nlp_(complex)ErbB2/ErbB3/neuregulin_1_beta_(complex) Nlp-depleted_centrosome_(complex) centrosome_(complex) GLI3

CSNK1D

Su(fu)/SIN3/HDAC_complex_(complex)GLI3/Su(fu)_(complex)

PIK3CD1 KCNJ11 FKBP5 LRP2 WNT7B STAT5B SUFU

Figure B.6: Pathways signiﬁcantly over-expressed in DTEPs are highly interconnected. Snapshot of Cytoscape [29, 119, 125, 132] visualization of the pathway connections between all “core” genes in pathways GSEA [138] found to be signiﬁcantly (FDR q-value < 0.1) over-expressed in DTEPs compared to untreated cells. “Core” genes are those that contribute to the GSEA enrichment score, and are therefore over- expressed in DTEPs. Gene nodes were colored based on the DESeq [3] assigned log fold changes and sized based on the DESeq assigned p-value. All colored genes have a DE-Seq Benjamini-Hochberg adjusted p-value less than 0.1. Connections between genes were determined using a superimposed pathway, which incorporates pathway information from the National Cancer Institute Pathway Interaction Database, Biocarta, Reactome and KEGG [60, 122, 91, 32]. Green nodes are over-expressed in DTEPs and purple nodes are over-expressed in untreated cells. Solid lines represent transcriptional regulation, dashed lines represent post-transcriptional regulation, circular nodes represent proteins, triangles represent families, rounded rectangles represent processes and hexagons represent complexes. Cytoscape session available in supplementary data C.6.

104 7&*$EUHDVWLQYDVLYHFDUFLQRPD %5&$ JHQHH[SUHVVLRQE\51$VHT ,OOXPLQD+L6HT $GG'DWDVHWV TCGA breast invasive carcinoma gene expression by RNASeq 

%5&$JHQHH[SUHVVLRQ ,OOXPLQD+L6HT

HER2 Luminal Luminal A Luminal B Luminal Basal

-3 PAM50 Subtype Jag1 Jag2 MYC Subgroup DLL1 DLL3 DLL4 DTX2 HES1 LFNG SSPO RFNG EP300 FURIN GATA3 CTBP2 NEURL MAML1 NUMBL NCOR2 CREBBP

Figure B.7: Jag1 and Jag2 are under-expressed in HER2+ breast cancer patients. UCSC Cancer Browser view of RNA-Seq gene expression from TCGA breast cancer patients for all 5 Notch ligands and other Notch pathway genes sorted by PAM50 assigned subtype [90, 45, 162]. Samples are grouped into 2 subgroups, HER2 subtype (red) and Luminal A, Luminal B and Basal subtypes (green), and compared using a t-test. Genes signiﬁcantly under-expressed in the HER2+ patients are highlighted in green. Similar results are obtained if Bonferroni correction is used.

105 Up in DTPs and DTEPs Up in DTEPs Up in DTEPs vs. vs. vs. untreated untreated untreated and DTPs IGF-1R IGF1 IGF1 IGFBP5 IGF-1R IGF-2R KDM4B IGF-2R IGFBP2 IGFBP2 IGF2BP3 IGFBP4 KDM4B IGFBP5 KDM6B IGF2BP3 KDM4B KDM6B

Table B.1: DE-Seq results conﬁrm that IGF-1R and KDM pathways are over- expressed in resistant cells. List of genes in IGF pathway and KDM protein family that are over-expressed in DTPs and DTEPs compared to untreated, over-expressed in DTEPs only compared to untreated or over-expressed in DTEPs compared to both untreated and DTPs. All genes listed have a DE-Seq [3] assigned Benjamini-Hochberg adjusted p-value below 0.1 and a fold change greater than 1.5.

106 !"#"$%&'%()*$ "*!+ ,-+"./ 0*!1 !"##$%&'%(*.$ #23 ,-+4 0*-$ !"#-5%&'%"#*. 6-+* ,789 &:;<'%=>:@%A,3-4B !"7$ 6*""5 C,3 .)-+* !C3 6*""/ *!+4 .)-+*! !7C 6*""D C*!1 .)-+*" !E7 6*##$ *!14 .,/C4 "!) 6*C (!.C4 .,C/#! "!C 61*$ (#7$ .360 "!E +-+ (6)$5 *21 "#785 +-+* (6C$ 1(!# "#78F +7,.%13!35 (63 3-+G* "#78E H-+ (7H$ 3,(.$ ",( ,-+$ (1H5 E,!. ",*#5 ,-+$* (32* G$8,>: ",*#/ ,-+5 06C5

Table B.2: Genes known to contribute to drug resistance. List of all genes found to be involved in drug resistance [52, 12, 47].

107 Appendix C

List of supplementary data ﬁles

1. Chapter 2 Supplementary Table S1: BamBam output of peaks and rearrangements

(TCGA-GBM Exome analysis)

2. Chapter 2 Supplementary Table S3: Data-based validation of exome analysis

(Whole Genome Sequencing analysis)

3. Chapter 3 Supplementary Table S5: DNA rearrangements and potential double

minute chromosomes/breakpoint-enriched regions

4. Chapter 4 Supplementary Table S2: Expressed mutations in DTEPs called by

RADIA

5. Chapter 4 Supplementary Table S4: DE-Seq results: list of diﬀerentially expressed

108 genes

6. Chapter 4 supplementary data: Pathways signiﬁcantly over-expressed in DTEPs

(Cytoscape ﬁle)

109 Bibliography

[1] K Aldape, M L Simmons, R L Davis, R Miike, J Wiencke, G Barger, M Lee, P Chen, and M Wrensch. Discrepancies in diagnoses of neuroepithelial neoplasms: the san francisco bay area adult glioma study. Cancer, 88(10):2342–9, May 2000.

[2] Kenneth Aldape, Peter C Burger, and Arie Perry. Clinicopathologic aspects of 1p/19q loss and the diagnosis of oligodendroglioma. Arch Pathol Lab Med, 131(2):242–51, Feb 2007.

[3] S. Anders and W. Huber. Diﬀerential expression analysis for sequence count data. Genome Biol, 11(10):R106, 2010.

[4] Christina L Appin and Daniel J Brat. Molecular genetics of gliomas. Cancer J, 20(1):66–72, 2014.

[5] A. C. Baege, G. L. Disbrow, and R. Schlegel. Igfbp-3, a marker of cellular senes- cence, is overexpressed in human papillomavirus-immortalized cervical cells and enhances igf-1-induced mitogenesis. J Virol, 78(11):5720–7, 2004.

[6] G Balaban-Malenbaum and F Gilbert. Double minute chromosomes and the homogeneously staining regions in chromosomes of a human neuroblastoma cell line. Science, 198(4318):739–41, Nov 1977.

[7] Leonora Balaj, Ryan Lessard, Lixin Dai, Yoon-Jae Cho, Scott L Pomeroy, Xan- dra O Breakeﬁeld, and Johan Skog. Tumour microvesicles contain retrotransposon elements and ampliﬁed oncogene sequences. Nat Commun, 2:180, 2011.

[8] P E Barker. Double minutes in human tumor cells. Cancer Genet Cytogenet, 5(1):81–94, Feb 1982.

[9] Adam J Bass, Michael S Lawrence, Lear E Brace, Alex H Ramos, Yotam Drier, Kristian Cibulskis, Carrie Sougnez, Douglas Voet, Gordon Saksena, Andrey

110 Sivachenko, Rui Jing, Melissa Parkin, Trevor Pugh, Roel G Verhaak, Nicolas Stransky, Adam T Boutin, Jordi Barretina, David B Solit, Evi Vakiani, Wenlin Shao, Yuji Mishina, Markus Warmuth, Jose Jimenez, Derek Y Chiang, Sabina Signoretti, William G Kaelin, Nicole Spardy, William C Hahn, Yujin Hoshida, Shuji Ogino, Ronald A Depinho, Lynda Chin, Levi A Garraway, Charles S Fuchs, Jose Baselga, Josep Tabernero, Stacey Gabriel, Eric S Lander, Gad Getz, and Matthew Meyerson. Genomic sequencing of colorectal adenocarcinomas identiﬁes a recurrent vti1a-tcf7l2 fusion. Nat Genet, 43(10):964–8, Oct 2011.

[10] Chetan Bettegowda, Nishant Agrawal, Yuchen Jiao, Mark Sausen, Laura D Wood, Ralph H Hruban, Fausto J Rodriguez, Daniel P Cahill, Roger McLendon, Gre- gory Riggins, Victor E Velculescu, Sueli Mieko Oba-Shinjo, Suely Kazue Naga- hashi Marie, Bert Vogelstein, Darell Bigner, Hai Yan, Nickolas Papadopoulos, and Kenneth W Kinzler. Mutations in cic and fubp1 contribute to human oligodendroglioma. Science, 333(6048):1453–5, Sep 2011.

[11] S H Bigner, J Mark, and D D Bigner. Cytogenetics of human brain tumors. Cancer Genet Cytogenet, 47(2):141–54, Jul 1990.

[12] B. G. Blair, A. Bardelli, and B. H. Park. Somatic alterations as the basis for resistance to targeted therapies. J Pathol, 232(2):244–54, 2014.

[13] L. P. Blair, J. Cao, M. R. Zou, J. Sayegh, and Q. Yan. Epigenetic regulation by lysine demethylase 5 (kdm5) enzymes in cancer. Cancers (Basel), 3(1):1383–404, 2011.

[14] T David Bourne and David Schiﬀ. Update on molecular ﬁndings, management and outcome in low-grade gliomas. Nat Rev Neurol, 6(12):695–701, Dec 2010.

[15] Cameron W Brennan, Roel G W Verhaak, Aaron McKenna, Benito Cam- pos, Houtan Noushmehr, Soﬁe R Salama, Siyuan Zheng, Debyani Chakravarty, J Zachary Sanborn, Samuel H Berman, Rameen Beroukhim, Brady Bernard, Chang-Jiun Wu, Giannicola Genovese, Ilya Shmulevich, Jill Barnholtz-Sloan, Lihua Zou, Rahulsimham Vegesna, Sachet A Shukla, Giovanni Ciriello, W K Yung, Wei Zhang, Carrie Sougnez, Tom Mikkelsen, Kenneth Aldape, Darell D Bigner, Erwin G Van Meir, Michael Prados, Andrew Sloan, Keith L Black, Jennifer Eschbacher, Gaetano Finocchiaro, William Friedman, David W An- drews, Abhijit Guha, Mary Iacocca, Brian P O’Neill, Greg Foltz, Jerome My- ers, Daniel J Weisenberger, Robert Penny, Raju Kucherlapati, Charles M Perou, D Neil Hayes, Richard Gibbs, Marco Marra, Gordon B Mills, Eric Lander, Paul Spellman, Richard Wilson, Chris Sander, John Weinstein, Matthew Meyerson, Stacey Gabriel, Peter W Laird, David Haussler, Gad Getz, Lynda Chin, and

111 TCGA Research Network. The somatic genomic landscape of glioblastoma. Cell, 155(2):462–77, Oct 2013.

[16] R Brookwell and F A Hunt. Formation of double minutes by breakdown of a homogeneously staining region in a refractory anemia with excess blasts. Cancer Genet Cytogenet, 34(1):47–52, Aug 1988.

[17] Gregory Cairncross, Meihua Wang, Edward Shaw, Robert Jenkins, David Brach- man, Jan Buckner, Karen Fink, Luis Souhami, Normand Laperriere, Walter Cur- ran, and Minesh Mehta. Phase iii trial of chemoradiotherapy for anaplastic oligodendroglioma: long-term results of rtog 9402. J Clin Oncol, 31(3):337–43, Jan 2013.

[18] J G Cairncross and D R Macdonald. Oligodendroglioma: a new chemosensitive tumor. J Clin Oncol, 8(12):2090–1, Dec 1990.

[19] J G Cairncross, K Ueki, M C Zlatescu, D K Lisle, D M Finkelstein, R R Hammond, J S Silver, P C Stark, D R Macdonald, Y Ino, D A Ramsay, and D N Louis. Speciﬁc genetic predictors of chemotherapeutic response and survival in patients with anaplastic oligodendrogliomas. J Natl Cancer Inst, 90(19):1473–9, Oct 1998.

[20] J Gregory Cairncross, Meihua Wang, Robert B Jenkins, Edward G Shaw, Caterina Giannini, David G Brachman, Jan C Buckner, Karen L Fink, Luis Souhami, Normand J Laperriere, Jason T Huse, Minesh P Mehta, and Walter J Curran, Jr. Beneﬁt from procarbazine, lomustine, and vincristine in oligodendroglial tumors is associated with mutation of idh. J Clin Oncol, 32(8):783–90, Mar 2014.

[21] Cancer Genome Atlas Research Network. Comprehensive genomic characterization deﬁnes human glioblastoma genes and core pathways. Nature, 455(7216):1061–8, Oct 2008.

[22] Cancer Genome Atlas Research Network, John N Weinstein, Eric A Collisson, Gordon B Mills, Kenna R Mills Shaw, Brad A Ozenberger, Kyle Ellrott, Ilya Shmulevich, Chris Sander, and Joshua M Stuart. The cancer genome atlas pan- cancer analysis project. Nat Genet, 45(10):1113–20, Oct 2013.

[23] S. Cara and I. F. Tannock. Retreatment of patients with the same chemotherapy: implications for clinical mechanisms of drug resistance. Ann Oncol, 12(1):23–7, 2001.

[24] Ken Chen, John W Wallis, Michael D McLellan, David E Larson, Joelle M Kalicki, Craig S Pohl, Sean D McGrath, Michael C Wendl, Qunyuan Zhang, Devin P Locke, Xiaoqi Shi, Robert S Fulton, Timothy J Ley, Richard K Wilson, Li Ding,

112 and Elaine R Mardis. Breakdancer: an algorithm for high-resolution mapping of genomic structural variation. Nat Methods, 6(9):677–81, Sep 2009.

[25] E. Chinchar, K. L. Makey, J. Gibson, F. Chen, S. A. Cole, G. C. Megason, S. Vi- jayakumar, L. Miele, and J. W. Gu. Sunitinib signiﬁcantly suppresses the proliferation, migration, apoptosis resistance, tumor angiogenesis and growth of triple- negative breast cancers but increases breast cancer stem cells. Vasc Cell, 6:12, 2014.

[26] Jeonghee Cho, Sandra Pastorino, Qing Zeng, Xiaoyin Xu, William Johnson, Scott Vandenberg, Roel Verhaak, Andrew D Cherniack, Hideo Watanabe, Amit Dutt, Jihyun Kwon, Ying S Chao, Robert C Onofrio, Derek Chiang, Yuki Yuza, Santosh Kesari, and Matthew Meyerson. Glioblastoma-derived epidermal growth factor receptor carboxyl-terminal deletion mutants are transforming and are sensitive to egfr-directed therapies. Cancer Res, 71(24):7587–96, Dec 2011.

[27] Giovanni Ciriello, Martin L Miller, B¨ulent Arman Aksoy, Yasin Senbabaoglu, Nikolaus Schultz, and Chris Sander. Emerging landscape of oncogenic signatures across human cancers. Nat Genet, 45(10):1127–1133, Sep 2013.

[28] Ami Citri, Kochupurakkal Bose Skaria, and Yosef Yarden. The deaf and the dumb: the biology of erbb-2 and erbb-3. Exp Cell Res, 284(1):54–65, Mar 2003.

[29] M. S. Cline, M. Smoot, E. Cerami, A. Kuchinsky, N. Landys, C. Workman, R. Christmas, I. Avila-Campilo, M. Creech, B. Gross, K. Hanspers, R. Isser- lin, R. Kelley, S. Killcoyne, S. Lotia, S. Maere, J. Morris, K. Ono, V. Pavlovic, A. R. Pico, A. Vailaya, P. L. Wang, A. Adler, B. R. Conklin, L. Hood, M. Kuiper, C. Sander, I. Schmulevich, B. Schwikowski, G. J. Warner, T. Ideker, and G. D. Bader. Integration of biological networks and gene expression data using cytoscape. Nat Protoc, 2(10):2366–82, 2007.

[30] B. Cohen, M. Shimizu, J. Izrailit, N. F. Ng, Y. Buchman, J. G. Pan, J. Dering, and M. Reedijk. Cyclin d1 is a direct target of jag1-mediated notch signaling in breast cancer. Breast Cancer Res Treat, 123(1):113–24, 2010.

[31] S W Coons, P C Johnson, B W Scheithauer, A J Yates, and D K Pearl. Improv- ing diagnostic accuracy and interobserver concordance in the classiﬁcation and grading of primary gliomas. Cancer, 79(7):1381–93, Apr 1997.

[32] D. Croft, A. F. Mundo, R. Haw, M. Milacic, J. Weiser, G. Wu, M. Caudy, P. Gara- pati, M. Gillespie, M. R. Kamdar, B. Jassal, S. Jupe, L. Matthews, B. May, S. Palatnik, K. Rothfels, V. Shamovsky, H. Song, M. Williams, E. Birney, H. Her-

113 mjakob, L. Stein, and P. D’Eustachio. The reactome pathway knowledgebase. Nucleic Acids Res, 42(Database issue):D472–7, 2014.

[33] F. De Santa, V. Narang, Z. H. Yap, B. K. Tusi, T. Burgold, L. Austenaa, G. Bucci, M. Caganova, S. Notarbartolo, S. Casola, G. Testa, W. K. Sung, C. L. Wei, and G. Natoli. Jmjd3 contributes to the control of gene expression in lps-activated macrophages. EMBO J, 28(21):3341–52, 2009.

[34] F. De Santa, M. G. Totaro, E. Prosperini, S. Notarbartolo, G. Testa, and G. Na- toli. The histone h3 lysine-27 demethylase jmjd3 links inﬂammation to inhibition of polycomb-mediated gene silencing. Cell, 130(6):1083–94, 2007.

[35] Kathleen Deiteren, Dirk Hendriks, Simon Scharp´e,and Anne Marie Lambeir. Carboxypeptidase m: Multiple alliances and unknown partners. Clin Chim Acta, 399(1-2):24–39, Jan 2009.

[36] David M Duda, Daniel C Scott, Matthew F Calabrese, Erik S Zimmerman, Ning Zheng, and Brenda A Schulman. Structural regulation of cullin-ring ubiquitin ligase complexes. Curr Opin Struct Biol, 21(2):257–64, Apr 2011.

[37] R. Edgar, M. Domrachev, and A. E. Lash. Gene expression omnibus: Ncbi gene expression and hybridization array data repository. Nucleic Acids Res, 30(1):207– 10, 2002.

[38] S. Eliasz, S. Liang, Y. Chen, M. A. De Marco, O. Machek, S. Skucha, L. Miele, and M. Bocchetta. Notch-1 stimulates survival of lung adenocarcinoma cells during hypoxia by activating the igf-1r pathway. Oncogene, 29(17):2488–98, 2010.

[39] M. Ernst and G. A. Rodan. Increased activity of insulin-like growth factor (igf) in osteoblastic cells in the presence of growth hormone (gh): positive correlation with the presence of the gh-induced igf-binding protein bp-3. Endocrinology, 127(2):807–14, 1990.

[40] Centers for Disease Control and Prevention. Leading causes of death, 2010.

[41] S. A. Forbes, G. Bhamra, S. Bamford, E. Dawson, C. Kok, J. Clements, A. Men- zies, J. W. Teague, P. A. Futreal, and M. R. Stratton. The catalogue of somatic mutations in cancer (cosmic). Curr Protoc Hum Genet, Chapter 10:Unit 10.11, 2008.

[42] Kinisha Gala and Sarat Chandarlapaty. Molecular pathways: Her3 targeted therapy. Clin Cancer Res, 20(6):1410–6, Mar 2014.

114 [43] Charles E Geyer, John Forster, Deborah Lindquist, Stephen Chan, C Gilles Romieu, Tadeusz Pienkowski, Agnieszka Jagiello-Gruszfeld, John Crown, Arlene Chan, Bella Kaufman, Dimosthenis Skarlos, Mario Campone, Neville Davidson, Mark Berger, Cristina Oliva, Stephen D Rubin, Steven Stein, and David Cameron. Lapatinib plus capecitabine for her2-positive advanced breast cancer. N Engl J Med, 355(26):2733–2743, Dec 2006.

[44] Caterina Giannini, Peter C Burger, Brian A Berkey, J Gregory Cairncross, Robert B Jenkins, Minesh Mehta, Walter J Curran, and Ken Aldape. Anaplastic oligodendroglial tumors: reﬁning the correlation among histopathology, 1p 19q deletion and clinical outcome in intergroup radiation therapy oncology group trial 9402. Brain Pathol, 18(3):360–9, Jul 2008.

[45] M. Goldman, B. Craft, T. Swatloski, K. Ellrott, M. Cline, M. Diekhans, S. Ma, C. Wilks, J. Stuart, D. Haussler, and J. Zhu. The ucsc cancer genomics browser: update 2013. Nucleic Acids Res, 41(Database issue):D949–54, 2013.

[46] Thierry Gorlia, Jean-Yves Delattre, Alba A Brandes, Johan M Kros, Martin J B Taphoorn, Mathilde C M Kouwenhoven, H J J A Bernsen, Marc Fr´enay, Cees C Tijssen, Denis Lacombe, and Martin J van den Bent. New clinical, pathological and molecular prognostic models and calculators in patients with locally diagnosed anaplastic oligodendroglioma or oligoastrocytoma. a prognostic factor analysis of european organisation for research and treatment of cancer brain tumour group study 26951. Eur J Cancer, 49(16):3477–85, Nov 2013.

[47] M. M. Gottesman. Mechanisms of cancer drug resistance. Annu Rev Med, 53:615– 27, 2002.

[48] D Graus-Porta, R R Beerli, J M Daly, and N E Hynes. Erbb-2, the preferred heterodimerization partner of all erbb receptors, is a mediator of lateral signaling. EMBO J, 16(7):1647–1655, Apr 1997.

[49] Chris D Greenman, Erin D Pleasance, Scott Newman, Fengtang Yang, Beiyuan Fu, Serena Nik-Zainal, David Jones, King Wai Lau, Nigel Carter, Paul A W Ed- wards, P Andrew Futreal, Michael R Stratton, and Peter J Campbell. Estimation of rearrangement phylogeny for cancer genomes. Genome Res, 22(2):346–61, Feb 2012.

[50] Christian Hartmann, Bettina Hentschel, Wolfgang Wick, David Capper, J¨orgFels- berg, Matthias Simon, Manfred Westphal, Gabriele Schackert, Richard Meyer- mann, Torsten Pietsch, Guido Reifenberger, Michael Weller, Markus Loeﬄer, and Andreas von Deimling. Patients with idh1 wild type anaplastic astrocytomas exhibit worse prognosis than idh1-mutated glioblastomas, and idh1 mutation sta-

115 tus accounts for the unfavorable prognostic eﬀect of higher age: implications for classiﬁcation of gliomas. Acta Neuropathol, 120(6):707–18, Dec 2010.

[51] F. E. Henken, J. De-Castro Arce, F. R¨osl,L. Bosch, C. J. Meijer, P. J. Sni- jders, and R. D. Steenbergen. The functional role of notch signaling in hpv- mediated transformation is dose-dependent and linked to ap-1 alterations. Cell Oncol (Dordr), 35(2):77–84, 2012.

[52] C. Holohan, S. Van Schaeybroeck, D. B. Longley, and P. G. Johnston. Cancer drug resistance: an evolving paradigm. Nat Rev Cancer, 13(10):714–26, 2013.

[53] F. Hsu, W. J. Kent, H. Clawson, R. M. Kuhn, M. Diekhans, and D. Haussler. The ucsc known genes. Bioinformatics, 22(9):1036–46, 2006.

[54] da W Huang, B. T. Sherman, and R. A. Lempicki. Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res, 37(1):1–13, 2009.

[55] da W Huang, B. T. Sherman, and R. A. Lempicki. Systematic and integrative analysis of large gene lists using david bioinformatics resources. Nat Protoc, 4(1):44–57, 2009.

[56] N. E. Hynes and H. A. Lane. Erbb receptors and cancer: the complexity of targeted inhibitors. Nat Rev Cancer, 5(5):341–54, 2005.

[57] Yuchen Jiao, Patrick J Killela, Zachary J Reitman, Ahmed B Rasheed, Christo- pher M Heaphy, Roeland F de Wilde, Fausto J Rodriguez, Sergio Rosemberg, Sueli Mieko Oba-Shinjo, Suely Kazue Nagahashi Marie, Chetan Bettegowda, Nis- hant Agrawal, Eric Lipp, Christopher Pirozzi, Giselle Lopez, Yiping He, Henry Friedman, Allan H Friedman, Gregory J Riggins, Matthias Holdhoff, Peter Burger, Roger McLendon, Darell D Bigner, Bert Vogelstein, Alan K Meeker, Kenneth W Kinzler, Nickolas Papadopoulos, Luis A Diaz, and Hai Yan. Frequent atrx, cic, fubp1 and idh1 mutations refine the classification of malignant gliomas. Oncotar- get, 3(7):709–22, Jul 2012.

[58] Brett E Johnson, Tali Mazor, Chibo Hong, Michael Barnes, Koki Aihara, Cory Y McLean, Shaun D Fouse, Shogo Yamamoto, Hiroki Ueda, Kenji Tatsuno, Saurabh Asthana, Llewellyn E Jalbert, Sarah J Nelson, Andrew W Bollen, W Clay Gustafson, Elise Charron, William A Weiss, Ivan V Smirnov, Jun S Song, Adam B Olshen, Soonmee Cha, Yongjun Zhao, Richard A Moore, Andrew J Mungall, Steven J M Jones, Martin Hirst, Marco A Marra, Nobuhito Saito, Hiroyuki Abu- ratani, Akitake Mukasa, Mitchel S Berger, Susan M Chang, Barry S Taylor, and

116 Joseph F Costello. Mutational analysis reveals the origin and therapy-driven evo- lution of recurrent glioma. Science, 343(6167):189–93, Jan 2014.

[59] F E Jones and D F Stern. Expression of dominant-negative erbb2 in the mammary gland of transgenic mice reveals a role in lobuloalveolar development and lactation. Oncogene, 18(23):3481–3490, Jun 1999.

[60] M. Kanehisa and S. Goto. Kegg: kyoto encyclopedia of genes and genomes. Nucleic Acids Res, 28(1):27–30, 2000.

[61] Kasthuri Kannan, Akiko Inagaki, Joachim Silber, Daniel Gorovets, Jianan Zhang, Edward R Kastenhuber, Adriana Heguy, John H Petrini, Timothy A Chan, and Jason T Huse. Whole-exome sequencing identiﬁes atrx mutation as a key molecular determinant in lower-grade glioma. Oncotarget, 3(10):1194–203, Oct 2012.

[62] H. Kennecke, R. Yerushalmi, R. Woods, M. C. Cheang, D. Voduc, C. H. Speers, T. O. Nielsen, and K. Gelmon. Metastatic behavior of breast cancer subtypes. J Clin Oncol, 28(20):3271–7, 2010.

[63] W. J. Kent. Blat–the blast-like alignment tool. Genome Res, 12(4):656–64, 2002.

[64] Patrick J Killela, Christopher J Pirozzi, Zachary J Reitman, Sian Jones, B Ahmed Rasheed, Eric Lipp, Henry Friedman, Allan H Friedman, Yiping He, Roger E McLendon, Darell D Bigner, and Hai Yan. The genetic landscape of anaplastic astrocytoma. Oncotarget, 5(6):1452–7, Mar 2014.

[65] Patrick J Killela, Zachary J Reitman, Yuchen Jiao, Chetan Bettegowda, Nishant Agrawal, Luis A Diaz, Jr, Allan H Friedman, Henry Friedman, Gary L Gallia, Beppino C Giovanella, Arthur P Grollman, Tong-Chuan He, Yiping He, Ralph H Hruban, George I Jallo, Nils Mandahl, Alan K Meeker, Fredrik Mertens, George J Netto, B Ahmed Rasheed, Gregory J Riggins, Thomas A Rosenquist, Mark Schiﬀ- man, Ie-Ming Shih, Dan Theodorescu, Michael S Torbenson, Victor E Velculescu, Tian-Li Wang, Nicolas Wentzensen, Laura D Wood, Ming Zhang, Roger E McLen- don, Darell D Bigner, Kenneth W Kinzler, Bert Vogelstein, Nickolas Papadopou- los, and Hai Yan. Tert promoter mutations occur frequently in gliomas and a subset of tumors derived from cells with low rates of self-renewal. Proc Natl Acad Sci U S A, 110(15):6021–6, Apr 2013.

[66] T. Kurata, K. Tamura, H. Kaneda, T. Nogami, H. Uejima, G. Asai Go, K. Nak- agawa, and M. Fukuoka. Eﬀect of re-treatment with geﬁtinib (’iressa’, zd1839) after acquisition of resistance. Ann Oncol, 15(1):173–4, 2004.

[67] Fabien Kuttler and Sabine Mai. Formation of non-random extrachromosomal

117 elements during development, diﬀerentiation and oncogenesis. Semin Cancer Biol, 17(1):56–64, Feb 2007.

[68] B. Langmead and S. L. Salzberg. Fast gapped-read alignment with bowtie 2. Nat Methods, 9(4):357–9, 2012.

[69] Iris Lavon, Miri Refael, Bracha Zelikovitch, Edna Shalom, and Tali Siegal. Serum dna can deﬁne tumor-speciﬁc genetic and epigenetic markers in gliomas of various grades. Neuro Oncol, 12(2):173–80, Feb 2010.

[70] Michael S Lawrence, Petar Stojanov, Paz Polak, Gregory V Kryukov, Kristian Cibulskis, Andrey Sivachenko, Scott L Carter, Chip Stewart, Craig H Mermel, Steven A Roberts, Adam Kiezun, Peter S Hammerman, Aaron McKenna, Yotam Drier, Lihua Zou, Alex H Ramos, Trevor J Pugh, Nicolas Stransky, Elena Hel- man, Jaegil Kim, Carrie Sougnez, Lauren Ambrogio, Elizabeth Nickerson, Erica Sheﬂer, Maria L Cort´es,Daniel Auclair, Gordon Saksena, Douglas Voet, Michael Noble, Daniel DiCara, Pei Lin, Lee Lichtenstein, David I Heiman, Timothy Fen- nell, Marcin Imielinski, Bryan Hernandez, Eran Hodis, Sylvan Baca, Austin M Dulak, Jens Lohr, Dan-Avi Landau, Catherine J Wu, Jorge Melendez-Zajgla, Al- fredo Hidalgo-Miranda, Amnon Koren, Steven A McCarroll, Jaume Mora, Ryan S Lee, Brian Crompton, Robert Onofrio, Melissa Parkin, Wendy Winckler, Kristin Ardlie, Stacey B Gabriel, Charles W M Roberts, Jaclyn A Biegel, Kimberly Stegmaier, Adam J Bass, Levi A Garraway, Matthew Meyerson, Todd R Golub, Dmitry A Gordenin, Shamil Sunyaev, Eric S Lander, and Gad Getz. Mutational heterogeneity in cancer and the search for new cancer-associated genes. Nature, 499(7457):214–8, Jul 2013.

[71] Rebecca J Leary, Mark Sausen, Isaac Kinde, Nickolas Papadopoulos, John D Carpten, David Craig, Joyce O’Shaughnessy, Kenneth W Kinzler, Giovanni Parmigiani, Bert Vogelstein, Luis A Diaz, Jr, and Victor E Velculescu. Detection of chromosomal alterations in the circulation of cancer patients with whole-genome sequencing. Sci Transl Med, 4(162):162ra154, Nov 2012.

[72] D. A. Lewis and D. F. Spandau. Uvb-induced activation of nf-kappab is regulated by the igf-1r and dependent on p38 mapk. J Invest Dermatol, 128(4):1022–9, 2008.

[73] B. Li and C. N. Dewey. Rsem: accurate transcript quantiﬁcation from rna-seq data with or without a reference genome. BMC Bioinformatics, 12:323, 2011.

[74] H. Li, B. Handsaker, A. Wysoker, T. Fennell, J. Ruan, N. Homer, G. Marth, G. Abecasis, R. Durbin, and 1000 Genome Project Data Processing Subgroup.

118 The sequence alignment/map format and samtools. Bioinformatics, 25(16):2078– 9, 2009.

[75] Heng Li and Richard Durbin. Fast and accurate short read alignment with burrows-wheeler transform. Bioinformatics, 25(14):1754–60, Jul 2009.

[76] Heng Li, Bob Handsaker, Alec Wysoker, Tim Fennell, Jue Ruan, Nils Homer, Ga- bor Marth, Goncalo Abecasis, Richard Durbin, and 1000 Genome Project Data Processing Subgroup. The sequence alignment/map format and samtools. Bioin- formatics, 25(16):2078–9, Aug 2009.

[77] Y. Lombardo, M. Faronato, A. Filipovic, V. Vircillo, L. Magnani, and R. C. Coombes. Nicastrin and notch4 drive endocrine therapy resistance and epithelial- mesenchymal transition in mcf7 breast cancer cells. Breast Cancer Res, 16(3):R62, 2014.

[78] David N Louis, Hiroko Ohgaki, Otmar D Wiestler, Webster K Cavenee, Pe- ter C Burger, Anne Jouvet, Bernd W Scheithauer, and Paul Kleihues. The 2007 who classiﬁcation of tumours of the central nervous system. Acta Neuropathol, 114(2):97–109, Aug 2007.

[79] David N Louis, Arie Perry, Peter Burger, David W Ellison, Guido Reifenberger, Andreas von Deimling, Kenneth Aldape, Daniel Brat, V Peter Collins, Charles Eberhart, Dominique Figarella-Branger, Gregory N Fuller, Felice Giangaspero, Caterina Giannini, Cynthia Hawkins, Paul Kleihues, Andrey Korshunov, Jo- han M Kros, M Beatriz Lopes, Ho-Keung Ng, Hiroko Ohgaki, Werner Paulus, Torsten Pietsch, Marc Rosenblum, Elisabeth Rushing, Figen Soylemezoglu, Otmar Wiestler, and Pieter Wesseling. International society of neuropathology-haarlem consensus guidelines for nervous system tumor classiﬁcation and grading. Brain Pathol, 24(5):429–35, Sep 2014.

[80] Y. Lu, X. Zi, Y. Zhao, D. Mascarenhas, and M. Pollak. Insulin-like growth factor-i receptor signaling and resistance to trastuzumab (herceptin). J Natl Cancer Inst, 93(24):1852–7, 2001.

[81] Gisela Lundberg, Anders H Rosengren, Ulf H˚akanson, Henrik Stewénius,Yuesh- eng Jin, Ylva Stewénius,Sven P˚ahlman,and David Gisselsson. Binomial mitotic segregation of mycn-carrying double minutes in neuroblastoma illustrates the role of randomness in oncogene amplification. PLoS One, 3(8):e3099, 2008.

[82] D R Macdonald, L E Gaspar, and J G Cairncross. Successful chemotherapy for newly diagnosed aggressive oligodendroglioma. Ann Neurol, 27(5):573–4, May 1990.

119 [83] L. Magnani, A. Stoeck, X. Zhang, A. L´anczky, A. C. Mirabella, T. L. Wang, B. Gyorﬀy, and M. Lupien. Genome-wide reprogramming of the chromatin landscape underlies endocrine therapy resistance in breast cancer. Proc Natl Acad Sci USA, 110(16):E1490–9, 2013.

[84] Marcel Martin, Lars Maßh¨ofer,Petra Temming, Sven Rahmann, Claudia Metz, Norbert Bornfeld, Johannes van de Nes, Ludger Klein-Hitpass, Alan G Hinneb- usch, Bernhard Horsthemke, Dietmar R Lohmann, and Michael Zeschnigk. Ex- ome sequencing identiﬁes recurrent somatic mutations in eif1ax and sf3b1 in uveal melanoma with disomy 3. Nat Genet, 45(8):933–6, Aug 2013.

[85] J Matthew McDonald, Siew Ju See, Ivo W Tremont, Howard Colman, Mark R Gilbert, Morris Groves, Peter C Burger, David N Louis, Caterina Giannini, Gre- gory Fuller, Sandra Passe, Hilary Blair, Robert B Jenkins, Helen Yang, Alicia Ledoux, Joann Aaron, Ulka Tipnis, Wei Zhang, Kenneth Hess, and Ken Aldape. The prognostic impact of histology and 1p/19q status in anaplastic oligodendroglial tumors. Cancer, 104(7):1468–77, Oct 2005.

[86] G B McGwire and R A Skidgel. Extracellular conversion of epidermal growth factor (egf) to des-arg53-egf by carboxypeptidase m. J Biol Chem, 270(29):17154– 8, Jul 1995.

[87] Andrew McPherson, Fereydoun Hormozdiari, Abdalnasser Zayed, Ryan Giuliany, Gavin Ha, Mark G F Sun, Malachi Griﬃth, Alireza Heravi Moussavi, Janine Senz, Nataliya Melnyk, Marina Pacheco, Marco A Marra, Martin Hirst, Torsten O Nielsen, S Cenk Sahinalp, David Huntsman, and Sohrab P Shah. defuse: an algorithm for gene fusion discovery in tumor rna-seq data. PLoS Comput Biol, 7(5):e1001138, May 2011.

[88] Mitch McVey and Sang Eun Lee. Mmej repair of double-strand breaks (director’s cut): deleted sequences and alternative endings. Trends Genet, 24(11):529–38, Nov 2008.

[89] D K Moscatello, M Holgado-Madruga, A K Godwin, G Ramirez, G Gunn, P W Zoltick, J A Biegel, R L Hayes, and A J Wong. Frequent expression of a mutant epidermal growth factor receptor in multiple human tumors. Cancer Res, 55(23):5536–9, Dec 1995.

[90] The Cancer Genome Atlas Network. Comprehensive molecular portraits of human breast tumours. Nature, 490(7418):61–70, 2012.

[91] The Cancer Genome Atlas Research Network. Comprehensive molecular characterization of clear cell renal cell carcinoma. Nature, 499(7456):43–9, 2013.

120 [92] Sam Ng, Eric A Collisson, Artem Sokolov, Theodore Goldstein, Abel Gonzalez- Perez, Nuria Lopez-Bigas, Christopher Benz, David Haussler, and Joshua M Stu- art. Paradigm-shift predicts the function of mutations in multiple cancers using pathway impact analysis. Bioinformatics, 28(18):i640–i646, Sep 2012.

[93] G. Nishibuchi, Y. Shibata, T. Hayakawa, N. Hayakawa, Y. Ohtani, K. Shinmyozu, H. Tagami, and J. I. Nakayama. Physical and functional interactions between the histone h3k4 demethylase kdm5a and the nucleosome remodeling and deacetylase (nurd) complex. J Biol Chem, 2014.

[94] Paul A Northcott, David J H Shih, John Peacock, Livia Garzia, A Sorana Morrissy, Thomas Zichner, Adrian M Stütz,Andrey Korshunov, JüriReimand, Steven E Schumacher, Rameen Beroukhim, David W Ellison, Christian R Mar- shall, Anath C Lionel, Stephen Mack, Adrian Dubuc, Yuan Yao, Vijay Ra- maswamy, Betty Luu, Adi Rolider, Florence M G Cavalli, Xin Wang, Marc Remke, Xiaochong Wu, Readman Y B Chiu, Andy Chu, Eric Chuah, Richard D Cor- bett, Gemma R Hoad, Shaun D Jackman, Yisu Li, Allan Lo, Karen L Mungall, Ka Ming Nip, Jenny Q Qian, Anthony G J Raymond, Nina T Thiessen, Richard J Varhol, Inanc Birol, Richard A Moore, Andrew J Mungall, Robert Holt, Daisuke Kawauchi, Martine F Roussel, Marcel Kool, David T W Jones, Hendrick Witt, Africa Fernandez-L, Anna M Kenney, Robert J Wechsler-Reya, Peter Dirks, Tzvi Aviv, Wieslawa A Grajkowska, Marta Perek-Polnik, Christine C Haberler, Olivier Delattre, StéphanieS Reynaud, Fran¸coisF Doz, Sarah S Pernet-Fattet, Byung- Kyu Cho, Seung-Ki Kim, Kyu-Chang Wang, Wolfram Scheurlen, Charles G Eber- hart, Michelle Fèvre-Montange, Anne Jouvet, Ian F Pollack, Xing Fan, Karin M Muraszko, G Yancey Gillespie, Concezio Di Rocco, Luca Massimi, Erna M C Michiels, Nanne K Kloosterhof, Pim J French, Johan M Kros, James M Olson, Richard G Ellenbogen, Karel Zitterbart, Leos Kren, Reid C Thompson, Michael K Cooper, Boleslaw Lach, Roger E McLendon, Darell D Bigner, Adam Fontebasso, Steffen Albrecht, Nada Jabado, Janet C Lindsey, Simon Bailey, Nalin Gupta, William A Weiss, LászlóBognár,Almos Klekner, Timothy E Van Meter, Toshihiro Kumabe, Teiji Tominaga, Samer K Elbabaa, Jeffrey R Leonard, Joshua B Rubin, Linda M Liau, Erwin G Van Meir, Maryam Fouladi, Hideo Nakamura, Giuseppe Cinalli, MiklósGarami, Peter Hauser, Ali G Saad, Achille Iolascon, Shin Jung, Carlos G Carlotti, Rajeev Vibhakar, Young Shin Ra, Shenandoah Robinson, Mas- simo Zollo, Claudia C Faria, Jennifer A Chan, Michael L Levy, Poul H B Sorensen, Matthew Meyerson, Scott L Pomeroy, Yoon-Jae Cho, Gary D Bader, Uri Tabori, Cynthia E Hawkins, Eric Bouffet, Stephen W Scherer, James T Rutka, David Malkin, Steven C Clifford, Steven J M Jones, Jan O Korbel, Stefan M Pfister, Marco A Marra, and Michael D Taylor. Subgroup-specific structural variation across 1,000 medulloblastoma genomes. Nature, 488(7409):49–56, Aug 2012.

[95] Houtan Noushmehr, Daniel J Weisenberger, Kristin Diefes, Heidi S Phillips,

121 Kanan Pujara, Benjamin P Berman, Fei Pan, Christopher E Pelloski, Erik P Sul- man, Krishna P Bhat, Roel G W Verhaak, Katherine A Hoadley, D Neil Hayes, Charles M Perou, Heather K Schmidt, Li Ding, Richard K Wilson, David Van Den Berg, Hui Shen, Henrik Bengtsson, Pierre Neuvial, Leslie M Cope, Jonathan Buckley, James G Herman, Stephen B Baylin, Peter W Laird, Kenneth Aldape, and Cancer Genome Atlas Research Network. Identiﬁcation of a cpg island methy- lator phenotype that deﬁnes a distinct subgroup of glioma. Cancer Cell, 17(5):510– 22, May 2010.

[96] R. Olsauskas-Kuprys, A. Zlobin, and C. Osipo. Gamma secretase inhibitors of notch signaling. Onco Targets Ther, 6:943–55, 2013.

[97] C. Osipo, P. Patel, P. Rizzo, A. G. Clementz, L. Hao, T. E. Golde, and L. Miele. Erbb-2 inhibition activates notch-1 and sensitizes breast cancer cells to a gamma- secretase inhibitor. Oncogene, 27(37):5019–32, 2008.

[98] Quinn T Ostrom, Haley Gittleman, Paul Farah, Annie Ondracek, Yanwen Chen, Yingli Wolinsky, Nancy E Stroup, Carol Kruchko, and Jill S Barnholtz-Sloan. Cbtrus statistical report: Primary brain and central nervous system tumors diagnosed in the united states in 2006-2010. Neuro Oncol, 15 Suppl 2:ii1–56, Nov 2013.

[99] T. Palomero, W. K. Lim, D. T. Odom, M. L. Sulis, P. J. Real, A. Margolin, K. C. Barnes, J. O’Neil, D. Neuberg, A. P. Weng, J. C. Aster, F. Sigaux, J. Soulier, A. T. Look, R. A. Young, A. Califano, and A. A. Ferrando. Notch1 directly regulates c- myc and activates a feed-forward-loop transcriptional network promoting leukemic cell growth. Proc Natl Acad Sci U S A, 103(48):18261–6, 2006.

[100] K. Pandya, K. Meeke, A. G. Clementz, A. Rogowski, J. Roberts, L. Miele, K. S. Albain, and C. Osipo. Targeting both notch and erbb-2 signalling pathways is required for prevention of erbb-2-positive breast tumour recurrence. Br J Cancer, 105(6):796–806, 2011.

[101] Brittany C Parker, Matti J Annala, David E Cogdell, Kirsi J Granberg, Yan Sun, Ping Ji, Xia Li, Joy Gumin, Hong Zheng, Limei Hu, Olli Yli-Harja, Hannu Haapasalo, Tapio Visakorpi, Xiuping Liu, Chang-Gong Liu, Raymond Sawaya, Gregory N Fuller, Kexin Chen, Frederick F Lang, Matti Nykter, and Wei Zhang. The tumorigenic fgfr3-tacc3 gene fusion escapes mir-99a regulation in glioblastoma. J Clin Invest, 123(2):855–65, Feb 2013.

[102] J. S. Parker, M. Mullins, M. C. Cheang, S. Leung, D. Voduc, T. Vickery, S. Davies, C. Fauron, X. He, Z. Hu, J. F. Quackenbush, I. J. Stijleman, J. Palazzo, J. S. Marron, A. B. Nobel, E. Mardis, T. O. Nielsen, M. J. Ellis, C. M. Perou, and P. S.

122 Bernard. Supervised risk predictor of breast cancer based on intrinsic subtypes. J Clin Oncol, 27(8):1160–7, 2009.

[103] Joel S Parker, Michael Mullins, Maggie C U Cheang, Samuel Leung, David Vo- duc, Tammi Vickery, Sherri Davies, Christiane Fauron, Xiaping He, Zhiyuan Hu, John F Quackenbush, Inge J Stijleman, Juan Palazzo, J S Marron, Andrew B Nobel, Elaine Mardis, Torsten O Nielsen, Matthew J Ellis, Charles M Perou, and Philip S Bernard. Supervised risk predictor of breast cancer based on intrinsic subtypes. J Clin Oncol, 27(8):1160–1167, Mar 2009.

[104] D. Parkhomchuk, T. Borodina, V. Amstislavskiy, M. Banaru, L. Hallen, S. Kro- bitsch, H. Lehrach, and A. Soldatov. Transcriptome analysis by strand-speciﬁc sequencing of complementary dna. Nucleic Acids Res, 37(18):e123, 2009.

[105] D Williams Parsons, SiˆanJones, Xiaosong Zhang, Jimmy Cheng-Ho Lin, Re- becca J Leary, Philipp Angenendt, Parminder Mankoo, Hannah Carter, I-Mei Siu, Gary L Gallia, Alessandro Olivi, Roger McLendon, B Ahmed Rasheed, Stephen Keir, Tatiana Nikolskaya, Yuri Nikolsky, Dana A Busam, Hanna Tekleab, Luis A Diaz, Jr, James Hartigan, Doug R Smith, Robert L Strausberg, Suely Kazue Na- gahashi Marie, Sueli Mieko Oba Shinjo, Hai Yan, Gregory J Riggins, Darell D Bigner, Rachel Karchin, Nick Papadopoulos, Giovanni Parmigiani, Bert Vogel- stein, Victor E Velculescu, and Kenneth W Kinzler. An integrated genomic analysis of human glioblastoma multiforme. Science, 321(5897):1807–12, Sep 2008.

[106] Carmen Phillips. Treatment options for her2-positive breast cancer expand and evolve, 2012.

[107] Heidi S Phillips, Samir Kharbanda, Ruihuan Chen, William F Forrest, Robert H Soriano, Thomas D Wu, Anjan Misra, Janice M Nigro, Howard Colman, Liliana Soroceanu, P Mickey Williams, Zora Modrusan, Burt G Feuerstein, and Ken Aldape. Molecular subclasses of high-grade glioma predict prognosis, delineate a pattern of disease progression, and resemble stages in neurogenesis. Cancer Cell, 9(3):157–73, Mar 2006.

[108] Sandhya Pruthi. Her2-positive breast cancer: What is it?, 2012.

[109] Aaron R Quinlan and Ira M Hall. Bedtools: a ﬂexible suite of utilities for comparing genomic features. Bioinformatics, 26(6):841–2, Mar 2010.

[110] Amie J Radenbaugh, Singer Ma, Adam Ewing, Joshua M Stuart, Eric A Collisson, Jingchun Zhu, and David Haussler. Radia: Rna and dna integrated analysis for somatic mutation detection. PLoS One, 9(11):e111516, 2014.

123 [111] Tobias Rausch, David T W Jones, Marc Zapatka, Adrian M Stütz,Thomas Zich- ner, Joachim Weischenfeldt, Natalie Jäger,Marc Remke, David Shih, Paul A Northcott, Elke Pfaff, Jelena Tica, Qi Wang, Luca Massimi, Hendrik Witt, Se- bastian Bender, Sabrina Pleier, Huriye Cin, Cynthia Hawkins, Christian Beck, Andreas von Deimling, Volkmar Hans, Benedikt Brors, Roland Eils, Wolfram Scheurlen, Jonathon Blake, Vladimir Benes, Andreas E Kulozik, Olaf Witt, Di- anna Martin, Cindy Zhang, Rinnat Porat, Diana M Merino, Jonathan Wasser- man, Nada Jabado, Adam Fontebasso, Lars Bullinger, Frank G Rücker, Kon- stanze Döhner,Hartmut Döhner,Jan Koster, Jan J Molenaar, Rogier Versteeg, Marcel Kool, Uri Tabori, David Malkin, Andrey Korshunov, Michael D Taylor, Peter Lichter, Stefan M Pfister, and Jan O Korbel. Genome sequencing of pe- diatric medulloblastoma links catastrophic dna rearrangements with tp53 mutations. Cell, 148(1-2):59–71, Jan 2012.

[112] K S Reddy. Double minutes (dmin) and homogeneously staining regions (hsr) in myeloid disorders: a new case suggesting that dmin form hsr in vivo. Cytogenet Genome Res, 119(1-2):53–9, 2007.

[113] M. Reedijk, D. Pinnaduwage, B. C. Dickson, A. M. Mulligan, H. Zhang, S. B. Bull, F. P. O’Malley, S. E. Egan, and I. L. Andrulis. Jag1 expression is associated with a basal phenotype and recurrence in lymph node-negative breast cancer. Breast Cancer Res Treat, 111(3):439–48, 2008.

[114] David Reverter, Klaus Maskos, Fulong Tan, Randal A Skidgel, and Wolfram Bode. Crystal structure of human carboxypeptidase m, a membrane-bound enzyme that regulates peptide hormone activity. J Mol Biol, 338(2):257–69, Apr 2004.

[115] Dan Rohle, Janeta Popovici-Muller, Nicolaos Palaskas, Sevin Turcan, Christian Grommes, Carl Campos, Jennifer Tsoi, Owen Clark, Barbara Oldrini, Evangelia Komisopoulou, Kaiko Kunii, Alicia Pedraza, Stefanie Schalm, Lee Silverman, Alexandra Miller, Fang Wang, Hua Yang, Yue Chen, Andrew Kernytsky, Marc K Rosenblum, Wei Liu, Scott A Biller, Shinsan M Su, Cameron W Brennan, Tim- othy A Chan, Thomas G Graeber, Katharine E Yen, and Ingo K Mellinghoﬀ. An inhibitor of mutant idh1 delays growth and promotes diﬀerentiation of glioma cells. Science, 340(6132):626–30, May 2013.

[116] R. Roskoski. The erbb/her family of protein-tyrosine kinases and cancer. Phar- macol Res, 79:34–74, 2014.

[117] S. Sahebjam, P. L. Bedard, V. Castonguay, Z. Chen, M. Reedijk, G. Liu, B. Cohen, W. J. Zhang, B. Clarke, T. Zhang, S. Kamel-Reid, H. Chen, S. P. Ivy, A. R. Razak, A. M. Oza, E. X. Chen, H. W. Hirte, A. McGarrity, L. Wang, L. L. Siu, and S. J.

124 Hotte. A phase i study of the combination of ro4929097 and cediranib in patients with advanced solid tumours (pjc-004/nci 8503). Br J Cancer, 109(4):943–9, 2013.

[118] Felix Sahm, David Reuss, Christian Koelsche, David Capper, Jens Schittenhelm, Stephanie Heim, David T W Jones, Stefan M Pﬁster, Christel Herold-Mende, Wolfgang Wick, Wolf Mueller, Christian Hartmann, Werner Paulus, and An- dreas von Deimling. Farewell to oligoastrocytoma: in situ molecular genetics favor classiﬁcation as either oligodendroglioma or astrocytoma. Acta Neuropathol, 128(4):551–9, Oct 2014.

[119] R. Saito, M. E. Smoot, K. Ono, J. Ruscheinski, P. L. Wang, S. Lotia, A. R. Pico, G. D. Bader, and T. Ideker. A travel guide to cytoscape plugins. Nat Methods, 9(11):1069–76, 2012.

[120] Nader Sanai, Susan Chang, and Mitchel S Berger. Low-grade gliomas in adults. J Neurosurg, 115(5):948–65, Nov 2011.

[121] J Zachary Sanborn, Soﬁe R Salama, Mia Griﬀord, Cameron W Brennan, Tom Mikkelsen, Suresh Jhanwar, Sol Katzman, Lynda Chin, and David Haussler. Dou- ble minute chromosomes in glioblastoma multiforme are revealed by precise reconstruction of oncogenic amplicons. Cancer Res, 73(19):6036–45, Oct 2013.

[122] C. F. Schaefer, K. Anthony, S. Krupa, J. Buchoﬀ, M. Day, T. Hannay, and K. H. Buetow. Pid: the pathway interaction database. Nucleic Acids Res, 37(Database issue):D674–9, 2009.

[123] C. A. Schneider, W. S. Rasband, and K. W. Eliceiri. Nih image to imagej: 25 years of image analysis. Nat Methods, 9(7):671–5, 2012.

[124] Heidi Schwarzenbach, Dave S B Hoon, and Klaus Pantel. Cell-free nucleic acids as biomarkers in cancer patients. Nat Rev Cancer, 11(6):426–37, Jun 2011.

[125] P. Shannon, A. Markiel, O. Ozier, N. S. Baliga, J. T. Wang, D. Ramage, N. Amin, B. Schwikowski, and T. Ideker. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res, 13(11):2498–504, 2003.

[126] S. V. Sharma, D. Y. Lee, B. Li, M. P. Quinlan, F. Takahashi, S. Maheswaran, U. McDermott, N. Azizian, L. Zou, M. A. Fischbach, K. K. Wong, K. Brandstetter, B. Wittner, S. Ramaswamy, M. Classon, and J. Settleman. A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations. Cell, 141(1):69–80, 2010.

[127] Suzanne Sindi, Elena Helman, Ali Bashir, and Benjamin J Raphael. A geometric

125 approach for classiﬁcation and comparison of structural variants. Bioinformatics, 25(12):i222–30, Jun 2009.

[128] Devendra Singh, Joseph Minhow Chan, Pietro Zoppoli, Francesco Niola, Ryan Sullivan, Angelica Castano, Eric Minwei Liu, Jonathan Reichel, Paola Porrati, Serena Pellegatta, Kunlong Qiu, Zhibo Gao, Michele Ceccarelli, Riccardo Ric- cardi, Daniel J Brat, Abhijit Guha, Ken Aldape, John G Golﬁnos, David Zagzag, Tom Mikkelsen, Gaetano Finocchiaro, Anna Lasorella, Raul Rabadan, and Anto- nio Iavarone. Transforming fusions of fgfr and tacc genes in human glioblastoma. Science, 337(6099):1231–5, Sep 2012.

[129] Johan Skog, Tom W¨urdinger,Sjoerd van Rijn, Dimphna H Meijer, Laura Gainche, Miguel Sena-Esteves, William T Curry, Jr, Bob S Carter, Anna M Krichevsky, and Xandra O Breakeﬁeld. Glioblastoma microvesicles transport rna and proteins that promote tumour growth and provide diagnostic biomarkers. Nat Cell Biol, 10(12):1470–6, Dec 2008.

[130] D J Slamon, G M Clark, S G Wong, W J Levin, A Ullrich, and W L McGuire. Human breast cancer: correlation of relapse and survival with ampliﬁcation of the her-2/neu oncogene. Science, 235(4785):177–182, Jan 1987.

[131] Green P Smit AFA, Hubley R. Repeatmasker open-3.0, 1996-2010.

[132] M. E. Smoot, K. Ono, J. Ruscheinski, P. L. Wang, and T. Ideker. Cytoscape 2.8: new features for data integration and network visualization. Bioinformatics, 27(3):431–2, 2011.

[133] American Cancer Society. Cancer facts & ﬁgures 2011. Atlanta: American Cancer Society, 2011.

[134] T Sorlie, C M Perou, R Tibshirani, T Aas, S Geisler, H Johnsen, T Hastie, M B Eisen, M van de Rijn, S S Jeﬀrey, T Thorsen, H Quist, J C Matese, P O Brown, D Botstein, P Eystein Lonning, and A L Borresen-Dale. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A, 98(19):10869–10874, Sep 2001.

[135] Philip J Stephens, Chris D Greenman, Beiyuan Fu, Fengtang Yang, Graham R Bignell, Laura J Mudie, Erin D Pleasance, King Wai Lau, David Beare, Lucy A Stebbings, Stuart McLaren, Meng-Lay Lin, David J McBride, Ignacio Varela, Ser- ena Nik-Zainal, Catherine Leroy, Mingming Jia, Andrew Menzies, Adam P Butler, Jon W Teague, Michael A Quail, John Burton, Harold Swerdlow, Nigel P Carter, Laura A Morsberger, Christine Iacobuzio-Donahue, George A Follows, Anthony R Green, Adrienne M Flanagan, Michael R Stratton, P Andrew Futreal, and Peter J

126 Campbell. Massive genomic rearrangement acquired in a single catastrophic event during cancer development. Cell, 144(1):27–40, Jan 2011.

[136] Clelia Tiziana Storlazzi, Angelo Lonoce, Maria C Guastadisegni, Domenico Trom- betta, Pietro D’Addabbo, Giulia Daniele, Alberto L’Abbate, Gemma Macchia, Cecilia Surace, Klaas Kok, Reinhard Ullmann, Stefania Purgato, Orazio Palumbo, Massimo Carella, Peter F Ambros, and Mariano Rocchi. Gene ampliﬁcation as double minutes or homogeneously staining regions in solid tumors: origin and structure. Genome Res, 20(9):1198–206, Sep 2010.

[137] B Streubel, P Valent, U Jäger,M Edelhäuser,H Wandt, T Wagner, T Büchner, K Lechner, and C Fonatsch. Amplification of the mll gene on double minutes, a homogeneously staining region, and ring chromosomes in five patients with acute myeloid leukemia or myelodysplastic syndrome. Genes Chromosomes Cancer, 27(4):380–6, Apr 2000.

[138] A. Subramanian, P. Tamayo, V. K. Mootha, S. Mukherjee, B. L. Ebert, M. A. Gillette, A. Paulovich, S. L. Pomeroy, T. R. Golub, E. S. Lander, and J. P. Mesirov. Gene set enrichment analysis: a knowledge-based approach for interpret- ing genome-wide expression proﬁles. Proc Natl Acad Sci U S A, 102(43):15545–50, 2005.

[139] Ghazaleh Tabatabai, Roger Stupp, Martin J van den Bent, Monika E Hegi, J¨orgC Tonn, Wolfgang Wick, and Michael Weller. Molecular diagnostics of gliomas: the clinical perspective. Acta Neuropathol, 120(5):585–92, Nov 2010.

[140] Brett J Theeler, W K Alfred Yung, Gregory N Fuller, and John F De Groot. Moving toward molecular classiﬁcation of diﬀuse gliomas in adults. Neurology, 79(18):1917–26, Oct 2012.

[141] Wandaliz Torres-Garc´ıa,Siyuan Zheng, Andrey Sivachenko, Rahulsimham Veg- esna, Qianghu Wang, Rong Yao, Michael F Berger, John N Weinstein, Gad Getz, and Roel G W Verhaak. Prada: pipeline for rna sequencing data analysis. Bioin- formatics, 30(15):2224–6, Aug 2014.

[142] Sevin Turcan, Armida W M Fabius, Alexandra Borodovsky, Alicia Pedraza, Cameron Brennan, Jason Huse, Agnes Viale, Gregory J Riggins, and Timothy A Chan. Eﬃcient induction of diﬀerentiation and growth inhibition in idh1 mutant glioma cells by the dnmt inhibitor decitabine. Oncotarget, 4(10):1729–36, Oct 2013.

[143] Martin J van den Bent. Interobserver variation of the histopathological diagnosis

127 in clinical trials on glioma: a clinician’s perspective. Acta Neuropathol, 120(3):297– 304, Sep 2010.

[144] P Varlet, A Jouvet, C Miquel, G Saint-Pierre, F Beuvon, and C Daumas-Duport. [criteria of diagnosis and grading of oligodendrogliomas or oligo-astrocytomas according to the who and sainte-anne classiﬁcations]. Neurochirurgie, 51(3-4 Pt 2):239–46, Sep 2005.

[145] Roel G W Verhaak, Katherine A Hoadley, Elizabeth Purdom, Victoria Wang, Yuan Qi, Matthew D Wilkerson, C Ryan Miller, Li Ding, Todd Golub, Jill P Mesirov, Gabriele Alexe, Michael Lawrence, Michael O’Kelly, Pablo Tamayo, Bar- bara A Weir, Stacey Gabriel, Wendy Winckler, Supriya Gupta, Lakshmi Jakkula, Heidi S Feiler, J Graeme Hodgson, C David James, Jann N Sarkaria, Cameron Brennan, Ari Kahn, Paul T Spellman, Richard K Wilson, Terence P Speed, Joe W Gray, Matthew Meyerson, Gad Getz, Charles M Perou, D Neil Hayes, and Cancer Genome Atlas Research Network. Integrated genomic analysis identiﬁes clinically relevant subtypes of glioblastoma characterized by abnormalities in pdgfra, idh1, egfr, and nf1. Cancer Cell, 17(1):98–110, Jan 2010.

[146] Nicolas Vogt, Sandrine-HélèneLefèvre,Fran¸coiseApiou, Anne-Marie Dutrillaux, Andrej Cör,Pascal Leuraud, Marie-France Poupon, Bernard Dutrillaux, Michelle Debatisse, and Bernard Malfoy. Molecular structure of double-minute chromosomes bearing amplified copies of the epidermal growth factor receptor gene in gliomas. Proc Natl Acad Sci U S A, 101(31):11368–73, Aug 2004.

[147] RA Weinberg. The Biology Of Cancer. Garland Science, Taylor & Francis Group, LLC, 2007.

[148] Michael Weller, Roger Stupp, Monika E Hegi, Martin van den Bent, Joerg C Tonn, Marc Sanson, Wolfgang Wick, and Guido Reifenberger. Personalized care in neuro-oncology coming of age: why we need mgmt and 1p/19q testing for malignant glioma patients in clinical practice. Neuro Oncol, 14 Suppl 4:iv100–8, Sep 2012.

[149] Patrick Y Wen and Santosh Kesari. Malignant gliomas in adults. N Engl J Med, 359(5):492–507, Jul 2008.

[150] Andrew P Weng, Adolfo A Ferrando, Woojoong Lee, John P Morris, 4th, Lewis B Silverman, Cheryll Sanchez-Irizarry, Stephen C Blacklow, A Thomas Look, and Jon C Aster. Activating mutations of notch1 in human t cell acute lymphoblastic leukemia. Science, 306(5694):269–71, Oct 2004.

[151] Benedikt Wiestler, David Capper, Tim Holland-Letz, Andrey Korshunov, An-

128 dreas von Deimling, Stefan Michael Pfister, Michael Platten, Michael Weller, and Wolfgang Wick. Atrx loss refines the classification of anaplastic gliomas and identifies a subgroup of idh mutant astrocytic tumors with better prognosis. Acta Neuropathol, 126(3):443–51, Sep 2013.

[152] Benedikt Wiestler, David Capper, Martin Sill, David T W Jones, Volker Hov- estadt, Dominik Sturm, Christian Koelsche, Anna Bertoni, Leonille Schweizer, Andrey Korshunov, Elisa K Weiß, Maximilian G Schliesser, Alexander Rad- bruch, Christel Herold-Mende, Patrick Roth, Andreas Unterberg, Christian Hart- mann, Torsten Pietsch, Guido Reifenberger, Peter Lichter, Bernhard Radlwim- mer, Michael Platten, Stefan M Pﬁster, Andreas von Deimling, Michael Weller, and Wolfgang Wick. Integrated dna methylation and copy-number proﬁling identify three clinically and biologically relevant groups of anaplastic glioma. Acta Neuropathol, 128(4):561–71, Oct 2014.

[153] M. D. Wilkerson and D. N. Hayes. Consensusclusterplus: a class discovery tool with conﬁdence assessments and item tracking. Bioinformatics, 26(12):1572–3, 2010.

[154] Wenle Xia, Robert J Mullin, Barry R Keith, Lei-Hua Liu, Hong Ma, David W Rusnak, Gary Owens, Krystal J Alligood, and Neil L Spector. Anti-tumor activity of gw572016: a dual tyrosine kinase inhibitor blocks egf activation of egfr/erbb2 and downstream erk1/2 and akt pathways. Oncogene, 21(41):6255–6263, Sep 2002.

[155] Hai Yan, D Williams Parsons, Genglin Jin, Roger McLendon, B Ahmed Rasheed, Weishi Yuan, Ivan Kos, Ines Batinic-Haberle, SiˆanJones, Gregory J Riggins, Henry Friedman, Allan Friedman, David Reardon, James Herndon, Kenneth W Kinzler, Victor E Velculescu, Bert Vogelstein, and Darell D Bigner. Idh1 and idh2 mutations in gliomas. N Engl J Med, 360(8):765–73, Feb 2009.

[156] S. Yano, E. Nakataki, S. Ohtsuka, M. Inayama, H. Tomimoto, N. Edakuni, S. Kak- iuchi, N. Nishikubo, H. Muguruma, and S. Sone. Retreatment of lung adenocarcinoma patients with geﬁtinib who had experienced favorable results from their initial treatment with this selective epidermal growth factor receptor inhibitor: a report of three cases. Oncol Res, 15(2):107–11, 2005.

[157] Christina Yau, Yixin Wang, Yi Zhang, John A Foekens, and Christopher C Benz. Young age, increased tumor proliferation and foxm1 expression predict early metastatic relapse only for endocrine-dependent breast cancers. Breast Cancer Res Treat, 126(3):803–810, Apr 2011.

[158] Stephen Yip, Yaron S Butterﬁeld, Olena Morozova, Suganthi Chittaranjan, Michael D Blough, Jianghong An, Inanc Birol, Charles Chesnelong, Readman

129 Chiu, Eric Chuah, Richard Corbett, Rod Docking, Marlo Firme, Martin Hirst, Shaun Jackman, Aly Karsan, Haiyan Li, David N Louis, Alexandra Maslova, Richard Moore, Annie Moradian, Karen L Mungall, Marco Perizzolo, Jenny Qian, Gloria Roldan, Eric E Smith, Jessica Tamura-Wells, Nina Thiessen, Richard Varhol, Samuel Weiss, Wei Wu, Sean Young, Yongjun Zhao, Andrew J Mungall, Steven J M Jones, Gregg B Morin, Jennifer A Chan, J Gregory Cairncross, and Marco A Marra. Concurrent cic mutations, idh mutations, and 1p/19q loss distinguish oligodendrogliomas from other cancers. J Pathol, 226(1):7–16, Jan 2012.

[159] M Yoshimoto, S R Caminada De Toledo, E M Monteiro Caran, M T de Seixas, M L de Martino Lee, S de Campos Vieira Abib, S M Vianna, S T Schettini, and J Anderson Duffles Andrade. Mycn gene amplification. identification of cell populations containing double minutes and homogeneously staining regions in neuroblastoma tumors. Am J Pathol, 155(5):1439–43, Nov 1999.

[160] J. Yun, A. Pannuti, I. Espinoza, H. Zhu, C. Hicks, X. Zhu, M. Caskey, P. Rizzo, G. D’Souza, K. Backus, M. F. Denning, J. Coon, M. Sun, E. H. Bresnick, C. Osipo, J. Wu, P. R. Strack, D. A. Tonetti, and L. Miele. Crosstalk between pkc and notch-4 in endocrine-resistant breast cancer cells. Oncogenesis, 2:e60, 2013.

[161] Siyuan Zheng, Jun Fu, Rahulsimham Vegesna, Yong Mao, Lindsey E Heath- cock, Wandaliz Torres-Garcia, Ravesanker Ezhilarasan, Shuzhen Wang, Aaron McKenna, Lynda Chin, Cameron W Brennan, W K Alfred Yung, John N Wein- stein, Kenneth D Aldape, Erik P Sulman, Ken Chen, Dimpy Koul, and Roel G W Verhaak. A survey of intragenic breakpoints in glioblastoma identiﬁes a distinct subset associated with poor survival. Genes Dev, 27(13):1462–72, Jul 2013.

[162] J. Zhu, J. Z. Sanborn, S. Benz, C. Szeto, F. Hsu, R. M. Kuhn, D. Karolchik, J. Archie, M. E. Lenburg, L. J. Esserman, W. J. Kent, D. Haussler, and T. Wang. The ucsc cancer genomics browser. Nat Methods, 6(4):239–40, 2009.

130